diff --git a/.github/pull_request_template.md b/.github/pull_request_template.md
index 2cfad8f..c263c96 100644
--- a/.github/pull_request_template.md
+++ b/.github/pull_request_template.md
@@ -19,4 +19,3 @@ Explain the change and why it’s needed.
 - [ ] Linted locally (`npm run lint`) and tests pass (`npm test`).
 - [ ] If user‑facing, commit message uses `feat:` or `fix:` (or `BREAKING CHANGE` where appropriate).
 - [ ] Updated docs if needed.
-
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 231a2d5..3a17012 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -2,9 +2,9 @@ name: CI
 
 on:
   push:
-    branches: [ main, master ]
+    branches: [main, master]
   pull_request:
-    branches: [ main, master ]
+    branches: [main, master]
 
 jobs:
   build:
diff --git a/.github/workflows/commitlint.yml b/.github/workflows/commitlint.yml
index 0e30bd0..ab53c79 100644
--- a/.github/workflows/commitlint.yml
+++ b/.github/workflows/commitlint.yml
@@ -19,4 +19,3 @@ jobs:
       - name: Lint commit messages
         run: |
           npx commitlint --from=${{ github.event.pull_request.base.sha }} --to=${{ github.sha }} --verbose
-
diff --git a/.husky/pre-push b/.husky/pre-push
index 70d3d75..b59b08a 100644
--- a/.husky/pre-push
+++ b/.husky/pre-push
@@ -4,4 +4,4 @@ if [ -n "$CI" ] || [ "$HUSKY" = "0" ] || [ -n "$HUSKY_SKIP_HOOKS" ]; then
   exit 0
 fi
 
-npm test && npm run build
+npm run lint --fix && npm run format:check && npm test && npm run build
diff --git a/.prettierrc.json b/.prettierrc.json
new file mode 100644
index 0000000..f8b0a39
--- /dev/null
+++ b/.prettierrc.json
@@ -0,0 +1,8 @@
+{
+  "printWidth": 100,
+  "singleQuote": true,
+  "semi": true,
+  "trailingComma": "es5",
+  "arrowParens": "always",
+  "tabWidth": 2
+}
diff --git a/.releaserc.json b/.releaserc.json
index 0400df1..29a946b 100644
--- a/.releaserc.json
+++ b/.releaserc.json
@@ -1,7 +1,5 @@
 {
-  "branches": [
-    "main"
-  ],
+  "branches": ["main"],
   "plugins": [
     ["@semantic-release/commit-analyzer", { "preset": "conventionalcommits" }],
     ["@semantic-release/release-notes-generator", { "preset": "conventionalcommits" }],
diff --git a/AGENTS.md b/AGENTS.md
index ff7d488..13d3554 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -1,6 +1,7 @@
 # Repository Guidelines
 
 ## Project Structure & Module Organization
+
 - `src/` TypeScript source (ESM). Entry: `src/index.ts` → `main.ts`; shared helpers in `src/shared/`.
 - `bin/` CLI launcher (`openrouter`) that loads `dist/index.js`.
 - `dist/` build output (generated by `npm run build`). Do not edit.
@@ -9,6 +10,7 @@
 - Config files: global `~/.config/openrouter-cli/config.json`; project overrides `.openrouterrc(.json|.yaml|.yml)`.
 
 ## Build, Test, and Development Commands
+
 - `npm run dev` — run CLI in TS directly (ts-node ESM). Example: `npm run dev -- ask "Hello"`.
 - `npm run build` — compile TypeScript to `dist/`.
 - `npm test` / `npm run test:watch` — run Vitest once / in watch mode.
@@ -17,6 +19,7 @@
 - After build: `openrouter --help` (via `bin/openrouter`) or `node dist/index.js`.
 
 ## Coding Style & Naming Conventions
+
 - TypeScript, ESM, 2-space indentation, single quotes allowed; prefer explicit return types for exported functions.
 - File names: lowercase (e.g., `main.ts`, `repl.ts`, `shared/openrouter.ts`).
 - ESM import paths in TS include `.js` (e.g., `import { x } from './main.js'`).
@@ -24,14 +27,17 @@
 - Linting via flat-config ESLint (`eslint.config.js`); pre-commit runs `lint-staged`.
 
 ## Testing Guidelines
+
 - Framework: Vitest. Place tests in `tests/*.spec.ts` with clear, isolated cases.
 - Cover config precedence, URL joining, and CLI option parsing. Example: see `tests/openrouter.spec.ts` for `joinUrl`.
 - Run `npm test` locally; keep tests deterministic (no network). Mock I/O when needed.
 
 ## Commit & Pull Request Guidelines
+
 - Follow Conventional Commits: `feat:`, `fix:`, `chore:`, `docs:`, `refactor:`, `test:`; use scopes when helpful (e.g., `feat(config): ...`). Use `npm run commit`.
 - PRs: concise description, linked issues, CLI examples (commands/output), and updated docs when behavior changes. Ensure `lint`, `test`, and `build` pass.
 
 ## Security & Configuration Tips
+
 - Prefer `OPENROUTER_API_KEY`/`OPENAI_API_KEY`; avoid committing secrets. Use `openrouter config --api-key` only if persistence is required.
 - Example: `openrouter config --model meta-llama/llama-3.1-8b-instruct --domain https://openrouter.ai/api/v1`.
diff --git a/CHANGELOG.md b/CHANGELOG.md
index e2fbe4d..f393012 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,17 +1,15 @@
 ## 1.0.0-beta.1 (2025-09-07)
 
-
 ### ⚠ BREAKING CHANGES
 
-* **story/iteration-one:** new features
+- **story/iteration-one:** new features
 
 ### Features
 
-* add initial project structure and configuration files ([03c77bd](https://github.com/jwill9999/openrouter-cli/commit/03c77bdb9eef1a5daa496383e6708625419fc6b2))
-* **story/iteration-one:** add husky, github workflows, releases and tags, npm release ([147b78c](https://github.com/jwill9999/openrouter-cli/commit/147b78cee9d56fd1d6cfe33e382db215e7b7e760))
-* **story/iteration-one:** fix release workflow ([ac42562](https://github.com/jwill9999/openrouter-cli/commit/ac42562cf525ae42e5804b25a080bf5bc877dfa0))
-
+- add initial project structure and configuration files ([03c77bd](https://github.com/jwill9999/openrouter-cli/commit/03c77bdb9eef1a5daa496383e6708625419fc6b2))
+- **story/iteration-one:** add husky, github workflows, releases and tags, npm release ([147b78c](https://github.com/jwill9999/openrouter-cli/commit/147b78cee9d56fd1d6cfe33e382db215e7b7e760))
+- **story/iteration-one:** fix release workflow ([ac42562](https://github.com/jwill9999/openrouter-cli/commit/ac42562cf525ae42e5804b25a080bf5bc877dfa0))
 
 ### Bug Fixes
 
-* **story/iteration-one:** Fix failing release ([496821b](https://github.com/jwill9999/openrouter-cli/commit/496821b94161f14f717b791dfd686f0cbe798479))
+- **story/iteration-one:** Fix failing release ([496821b](https://github.com/jwill9999/openrouter-cli/commit/496821b94161f14f717b791dfd686f0cbe798479))
diff --git a/README.md b/README.md
index 8cee9d4..6a235b0 100644
--- a/README.md
+++ b/README.md
@@ -3,49 +3,97 @@
 [![npm latest](https://img.shields.io/npm/v/@letuscode/openrouter-cli)](https://www.npmjs.com/package/@letuscode/openrouter-cli)
 [![npm beta](https://img.shields.io/npm/v/@letuscode/openrouter-cli/beta)](https://www.npmjs.com/package/@letuscode/openrouter-cli?activeTab=versions)
 
-OpenAI‑compatible CLI for OpenRouter. Ask questions, run a REPL, and manage per‑project or global settings.
+OpenAI‑compatible CLI for OpenRouter. Ask questions, chat in a REPL, and fuzzy‑search models.
 
-Requirements
-- Node.js 18.17+ (ESM)
+You can change your model any time. In a terminal, run `openrouter models` to browse, or in the REPL type `/model` to search inline. Tip: search for `free` to see free models.
+
+## Install
 
-Install
 - Global: `npm i -g @letuscode/openrouter-cli`
 - One‑off: `npx @letuscode/openrouter-cli --help`
 
-Quick start
-1) Run the wizard: `openrouter init` (select provider, set domain/model, and add an API key)
-2) Ask something: `openrouter ask --no-stream "Hello!"`
-3) Chat interactively: `openrouter repl`
-
-Core commands
-- `openrouter init` — interactive setup (provider, domain, key, model, profile)
-- `openrouter config` — view config or set API key
-  - Examples:
-    - `openrouter config --list`
-    - `openrouter config --api-key sk-...` (stores in base config)
-    - `openrouter config --profile dev --api-key sk-...` (stores in profile)
-- `openrouter test` — verify connectivity (`/models`)
-  - `openrouter test [--profile dev] [--no-init]`
-- `openrouter ask` — one‑shot prompt
-  - `openrouter ask "your question" [-s SYSTEM] [--format auto|plain|md] [--profile NAME] [--no-stream] [--no-init]`
+Tip: Running `openrouter` with no args starts the setup wizard and then opens the REPL (in a terminal).
+
+### Requirements
+
+- Node.js 18.17+ (ESM)
+
+## Quick start
+
+1. Create an API key: https://openrouter.ai/keys
+2. Run setup: `openrouter` (or `openrouter init`) — enter your key if asked, then pick a model
+3. Ask once: `openrouter ask "Hello!"` — formatted answer by default
+4. Chat: `openrouter repl` — formatted replies; toggle streaming when you like
+
+## Everyday commands
+
+- `openrouter` (or `openrouter init`) — setup; uses the OpenRouter domain automatically; asks for a key only if missing; lets you pick a model; opens the REPL afterwards
+- `openrouter ask "…"` — answer a single question (formatted by default)
 - `openrouter repl` — interactive chat
-  - REPL commands: `exit`, `/model <name>`, `/system <text>`, `/format md|plain`, `/stream on|off`
+  - In the REPL:
+    - `/model` → inline search; type a few letters, pick a match
+    - `/model <id>` → set a specific model
+    - `/format md|plain` → formatted or plain replies (non‑stream)
+    - `/stream on|off` → stream tokens or wait for a full reply
+    - `exit` → quit
+- `openrouter models [query]` — browse models (fuzzy search) in a terminal; prints a table in non‑TTY
+- `openrouter config --list` — show current settings (keys are masked)
+- `openrouter config --api-key sk-…` — set your key once (or use the env var below)
+
+## Behavior & defaults
+
+- Ask: non‑stream + markdown rendering by default. Add `--stream` to stream tokens.
+- REPL: streaming OFF by default; markdown rendering for full replies; inline `/model` search.
 
-Configuration
-- API key via env (recommended): `export OPENROUTER_API_KEY=...` (or `OPENAI_API_KEY`)
-- Global config file: `~/.config/openrouter-cli/config.json` (chmod 600 where possible; keys never logged)
+## Configuration
+
+- API key via env (recommended): `export OPENROUTER_API_KEY=…` (or `OPENAI_API_KEY`)
+- Global config file: `~/.config/openrouter-cli/config.json` (private; keys never logged)
 - Project overrides: add `.openrouterrc` (JSON or YAML) in your project root
   - Example `.openrouterrc` (JSON):
     {
-      "domain": "http://localhost:11434/v1",
-      "model": "gemma2:9b-instruct"
+    "domain": "https://openrouter.ai/api/v1",
+    "model": "openrouter/auto"
     }
-- Changing default provider, domain, or model: re‑run `openrouter init` (this is the only way to update these defaults).
-- More details: see `docs/CONFIGURATION.md`.
+- Domain: fixed to the OpenRouter domain today (no prompt); kept in config for future provider choices
+- Change your default model any time by running `openrouter` again
+- Precedence: project rc > profile > global; env keys override persisted keys
+
+## Model search
+
+- `openrouter models` opens an interactive search in a terminal (type 2–3 letters)
+- `openrouter models llama` starts with “llama” suggestions; prints a table in non‑TTY
+
+### Example: inline model search in REPL
+
+```text
+(openai/gpt-oss-20b:free) > /model
+Search models (>=2 chars, blank to cancel): free
+Matches:
+1. openai/gpt-oss-120b:free — OpenAI: gpt-oss-120b (free)
+2. openai/gpt-oss-20b:free — OpenAI: gpt-oss-20b (free)
+…
+Pick 1-10 or type a model id:
+```
+
+## Output & accessibility
+
+- Non‑stream answers render markdown (bold/italic, headings, lists, inline code). Streaming prints raw tokens for responsiveness.
+- A “Thinking” spinner shows while waiting; colors/spinners honor `NO_COLOR` and TTY detection.
+
+## Troubleshooting
+
+- Missing API key: set `OPENROUTER_API_KEY` or run `openrouter` again. View current config: `openrouter config --list`.
+- “Policy / free endpoints” error: open https://openrouter.ai/settings/privacy and enable free endpoints, or choose a different model (`openrouter models`).
+- Picker shows a table: run in a terminal (TTY). Check: `node -p "process.stdout.isTTY && process.stdin.isTTY"`.
+- Friendly errors are shown; details are logged to `~/.config/openrouter-cli/cli.log`.
+
+## Advanced flags (optional)
+
+- Ask: `--stream`, `--format auto|plain|md`, `-s, --system <text>`, `--profile <name>`, `--no-init`
+- Models: `--non-interactive`
+- Config (debug): `--danger-reset`, `--override-json '<json>'`
 
-Troubleshooting
-- “Missing API key”: set `OPENROUTER_API_KEY` or run `openrouter init` (or `openrouter config --api-key sk-...`).
-- Non‑TTY/CI: pass `--no-init` to skip interactive prompts.
+## License
 
-License
 - MIT
diff --git a/bin/openrouter b/bin/openrouter
index 883d8e1..b1d5c53 100755
--- a/bin/openrouter
+++ b/bin/openrouter
@@ -1,3 +1,2 @@
 #!/usr/bin/env node
 import('../dist/index.js');
-
diff --git a/commitlint.config.cjs b/commitlint.config.cjs
index 8d7637c..f545120 100644
--- a/commitlint.config.cjs
+++ b/commitlint.config.cjs
@@ -4,6 +4,6 @@ module.exports = {
     'subject-case': [0],
     // Allow long auto-generated release notes from semantic-release
     'body-max-line-length': [0],
-    'footer-max-line-length': [0]
+    'footer-max-line-length': [0],
   },
 };
diff --git a/docs/CODE_REVIEW.md b/docs/CODE_REVIEW.md
index 6c0de73..886e3fc 100644
--- a/docs/CODE_REVIEW.md
+++ b/docs/CODE_REVIEW.md
@@ -1,124 +1,83 @@
- Yes — Iteration 1 is complete. Here’s a focused code review with prioritized improvements based on best practices
-  and SOLID principles.
-
-  Top Priority (Correctness + UX)
-
-  - REPL conversation state
-      - Issue: REPL streams assistant output to stdout but does not add assistant messages to history; subsequent
-  turns only include prior user messages.
-      - Fix: Capture streamed text (aggregate deltas) and append as an assistant message to history after each
-  turn.
-      - Side benefit: Enables proper multi-turn context.
-      - Side benefit: Enables proper multi-turn context.
-  -
-  HTTP robustness
-      - Add request timeouts via AbortController to avoid hanging calls.
-      - Normalize network errors and surface actionable messages (timeout, 401/403, 429 with retry-after).
-      - Respect HTTP_PROXY/HTTPS_PROXY/NO_PROXY and allow --timeout flag.
-  -
-  CLI ergonomics
-      - ask should accept stdin when prompt arg is omitted: echo "hi" | openrouter ask.
-      - Add --verbose for debug info (domain/model, request id if available).
-      - Improve test output (summarize model count, maybe first 3 model IDs).
-
-  Security + Config
-
-  - Secrets handling
-      - Current: never logs API keys (good), redact in config (good).
-      - Improve: redact any error payload that might echo credentials (defense-in-depth).
-      - Option: add --no-store on config --api-key to avoid writing key to disk (env only).
-      - Option: add --no-store on config --api-key to avoid writing key to disk (env only).
-  -
-  Config precedence
-      - Document and enforce: CLI flags > env > global config (current behavior aligns, document it explicitly).
-      - Add openrouter config --unset <key> to remove persisted values.
-      - Validate --domain (must be http/https).
-  -
-  File permissions
-      - You set 600/700 best-effort (good). On non-POSIX systems, log a warning if chmod fails.
-
-  Structure + SOLID
-
-  - Single responsibility
-      - Split src/main.ts subcommands into files: commands/config.ts, commands/test.ts, commands/ask.ts, commands/
-  repl.ts to isolate logic and keep main.ts as composition root.
-      - Extract an SseStreamer utility for streaming parsing (reusable in REPL and ask).
-      - Extract an SseStreamer utility for streaming parsing (reusable in REPL and ask).
-  -
-  Abstractions
-      - Introduce a ChatClient class with clear interface: chat(messages, opts), stream(messages, onDelta, opts).
+Yes — Iteration 1 is complete. Here’s a focused code review with prioritized improvements based on best practices
+and SOLID principles.
+
+Top Priority (Correctness + UX)
+
+- REPL conversation state - Issue: REPL streams assistant output to stdout but does not add assistant messages to history; subsequent
+  turns only include prior user messages. - Fix: Capture streamed text (aggregate deltas) and append as an assistant message to history after each
+  turn. - Side benefit: Enables proper multi-turn context. - Side benefit: Enables proper multi-turn context.
+- HTTP robustness - Add request timeouts via AbortController to avoid hanging calls. - Normalize network errors and surface actionable messages (timeout, 401/403, 429 with retry-after). - Respect HTTP_PROXY/HTTPS_PROXY/NO_PROXY and allow --timeout flag.
+- CLI ergonomics - ask should accept stdin when prompt arg is omitted: echo "hi" | openrouter ask. - Add --verbose for debug info (domain/model, request id if available). - Improve test output (summarize model count, maybe first 3 model IDs).
+
+Security + Config
+
+- Secrets handling
+  - Current: never logs API keys (good), redact in config (good).
+  - Improve: redact any error payload that might echo credentials (defense-in-depth).
+  - Option: add --no-store on config --api-key to avoid writing key to disk (env only).
+  - Option: add --no-store on config --api-key to avoid writing key to disk (env only).
+- Config precedence - Document and enforce: CLI flags > env > global config (current behavior aligns, document it explicitly). - Add openrouter config --unset <key> to remove persisted values. - Validate --domain (must be http/https).
+- File permissions - You set 600/700 best-effort (good). On non-POSIX systems, log a warning if chmod fails.
+
+Structure + SOLID
+
+- Single responsibility - Split src/main.ts subcommands into files: commands/config.ts, commands/test.ts, commands/ask.ts, commands/
+  repl.ts to isolate logic and keep main.ts as composition root. - Extract an SseStreamer utility for streaming parsing (reusable in REPL and ask). - Extract an SseStreamer utility for streaming parsing (reusable in REPL and ask).
+- Abstractions - Introduce a ChatClient class with clear interface: chat(messages, opts), stream(messages, onDelta, opts).
   This aids testing and swapping backends later (OpenAI/Ollama/Anthropic).
-  -
-  Dependency inversion
-      - Pass fetch (and logger) as injectable dependencies where feasible to ease testing/mocking.
-
-  Networking + Streaming
-
-  - SSE parsing
-      - Current manual buffer parsing works but can be brittle. Consider eventsource-parser for resilient SSE
-  handling (when you’re ready to add deps).
-      - Ensure you handle multi-line data events; guard against partial frames across chunks.
-      - Ensure you handle multi-line data events; guard against partial frames across chunks.
-  -
-  Headers
-      - Optionally include X-Title and HTTP-Referer headers for OpenRouter’s dashboards if desired (non-breaking).
-
-  Type Safety + TS Hygiene
-
-  - Types
-      - Define and export explicit types for API responses (minimal shapes used), e.g., ChatCompletionChunk,
-  ModelList.
-      - Enable noUncheckedIndexedAccess (optional) to catch missing choices?.[0] workflows.
-      - Enable noUncheckedIndexedAccess (optional) to catch missing choices?.[0] workflows.
-  -
-  Lint/Format
-      - You’ve set ESLint flat config with TS + globals (good). Consider adding Prettier or eslint-plugin-format
+- Dependency inversion - Pass fetch (and logger) as injectable dependencies where feasible to ease testing/mocking.
+
+Networking + Streaming
+
+- SSE parsing - Current manual buffer parsing works but can be brittle. Consider eventsource-parser for resilient SSE
+  handling (when you’re ready to add deps). - Ensure you handle multi-line data events; guard against partial frames across chunks. - Ensure you handle multi-line data events; guard against partial frames across chunks.
+- Headers - Optionally include X-Title and HTTP-Referer headers for OpenRouter’s dashboards if desired (non-breaking).
+
+Type Safety + TS Hygiene
+
+- Types - Define and export explicit types for API responses (minimal shapes used), e.g., ChatCompletionChunk,
+  ModelList. - Enable noUncheckedIndexedAccess (optional) to catch missing choices?.[0] workflows. - Enable noUncheckedIndexedAccess (optional) to catch missing choices?.[0] workflows.
+- Lint/Format - You’ve set ESLint flat config with TS + globals (good). Consider adding Prettier or eslint-plugin-format
   later for consistent formatting.
 
-  Testing
+Testing
 
-  - Unit tests to add (high value, low effort)
-      - joinUrl(base, path) — path preservation (e.g., /api/v1 + chat/completions).
-      - Config I/O: read/write, redaction, chmod best-effort no-throw.
-      - Env precedence: env > config; masking behavior.
-      - Streaming parser: simulate SSE frames (with and without chunk boundaries).
-  - E2E smoke (later)
-      - CLI test mocked via Nock or MSW Node to avoid network dependency.
-      - ask --no-stream pipeline with mocked responses.
+- Unit tests to add (high value, low effort)
+  - joinUrl(base, path) — path preservation (e.g., /api/v1 + chat/completions).
+  - Config I/O: read/write, redaction, chmod best-effort no-throw.
+  - Env precedence: env > config; masking behavior.
+  - Streaming parser: simulate SSE frames (with and without chunk boundaries).
+- E2E smoke (later)
+  - CLI test mocked via Nock or MSW Node to avoid network dependency.
+  - ask --no-stream pipeline with mocked responses.
 
-  CI/CD
+CI/CD
 
-  - CI install step
-      - In release.yml, replace npm ci with npm install --no-audit --no-fund (no lockfile checked in). You already
-  use npm install elsewhere; make it consistent to avoid failures.
-      - In release.yml, replace npm ci with npm install --no-audit --no-fund (no lockfile checked in). You already
+- CI install step - In release.yml, replace npm ci with npm install --no-audit --no-fund (no lockfile checked in). You already
+  use npm install elsewhere; make it consistent to avoid failures. - In release.yml, replace npm ci with npm install --no-audit --no-fund (no lockfile checked in). You already
   use npm install elsewhere; make it consistent to avoid failures.
-  -
-  Beta guarding (optional)
-      - Current beta publishes on every push (now safe-guarded from duplicates). Consider gating publish to only
+- Beta guarding (optional) - Current beta publishes on every push (now safe-guarded from duplicates). Consider gating publish to only
   run when commits include feat/fix/perf or a label like publish-beta.
-  -
-  Provenance
-      - You’ve set id-token and provenance (good). Add NPM_CONFIG_PROVENANCE=true in the release step env if you
+- Provenance - You’ve set id-token and provenance (good). Add NPM_CONFIG_PROVENANCE=true in the release step env if you
   want to ensure provenance for stable too.
 
-  Docs
+Docs
 
-  - README
-      - Document precedence order clearly.
-      - Add stdin usage example for ask.
-      - Add REPL commands table and mention multi-turn context.
+- README
+  - Document precedence order clearly.
+  - Add stdin usage example for ask.
+  - Add REPL commands table and mention multi-turn context.
 
-  Package metadata
+Package metadata
 
-  - Author field is empty; consider filling author and contributors for clarity.
-  - Consider preferGlobal: true to hint it’s a CLI.
+- Author field is empty; consider filling author and contributors for clarity.
+- Consider preferGlobal: true to hint it’s a CLI.
 
-  Misc UX
+Misc UX
 
-  - openrouter config default behavior shows current config (good). Add --json (explicit) and a terser human-
+- openrouter config default behavior shows current config (good). Add --json (explicit) and a terser human-
   readable output by default (optional).
-  - repl command: add /help to list commands; echo current model/system on start.
+- repl command: add /help to list commands; echo current model/system on start.
 
-  If you want, I can implement a first pass of the critical fixes (REPL history capture, stdin for ask, timeout
-  support, CI npm install change) in a PR-sized patch and add a couple of unit tests for joinUrl and config I/O.
\ No newline at end of file
+If you want, I can implement a first pass of the critical fixes (REPL history capture, stdin for ask, timeout
+support, CI npm install change) in a PR-sized patch and add a couple of unit tests for joinUrl and config I/O.
diff --git a/docs/CONFIGURATION.md b/docs/CONFIGURATION.md
index e4a769d..1253b20 100644
--- a/docs/CONFIGURATION.md
+++ b/docs/CONFIGURATION.md
@@ -6,18 +6,20 @@ This guide explains how configuration works in the CLI, how to set API keys secu
 
 Configuration comes from multiple sources. At runtime, values are resolved using this precedence (highest → lowest):
 
-1) CLI flags (e.g., `--profile`, `-m/--model`, `--domain`)
-2) Environment variables (API key: `OPENROUTER_API_KEY` or `OPENAI_API_KEY`)
-3) Project `.openrouterrc` (JSON or YAML in your project root)
-4) Global profile (`profiles.<name>` in `~/.config/openrouter-cli/config.json`)
-5) Global base config (`~/.config/openrouter-cli/config.json`)
-6) Built‑in defaults
+1. CLI flags (e.g., `--profile`, `-m/--model`, `--domain`)
+2. Environment variables (API key: `OPENROUTER_API_KEY` or `OPENAI_API_KEY`)
+3. Project `.openrouterrc` (JSON or YAML in your project root)
+4. Global profile (`profiles.<name>` in `~/.config/openrouter-cli/config.json`)
+5. Global base config (`~/.config/openrouter-cli/config.json`)
+6. Built‑in defaults
 
 Defaults:
+
 - Domain: `https://openrouter.ai/api/v1`
 - Model: `meta-llama/llama-3.1-8b-instruct`
 
 Interactive onboarding:
+
 - Run `openrouter init` to select a provider preset, set domain/model, and (optionally) persist an API key. Commands like `ask`, `test`, and `repl` auto‑prompt on missing keys in TTY unless `--no-init` is provided.
 
 ## Global Config
@@ -25,6 +27,7 @@ Interactive onboarding:
 - Path: `~/.config/openrouter-cli/config.json`
 - Stores base values and optional `profiles` object.
 - Example:
+
 ```json
 {
   "domain": "https://openrouter.ai/api/v1",
@@ -37,6 +40,7 @@ Interactive onboarding:
   }
 }
 ```
+
 - Permissions: written with chmod 600 where possible. Keys are redacted when printed.
 
 ### Managing Global Config via CLI
@@ -56,13 +60,16 @@ Interactive onboarding:
 Place a `.openrouterrc` in your project root to set project‑specific defaults. This file should not contain secrets.
 
 - JSON example (`./.openrouterrc` or `./.openrouterrc.json`):
+
 ```json
 {
   "domain": "http://localhost:11434/v1",
   "model": "gemma2:9b-instruct"
 }
 ```
+
 - YAML example (`./.openrouterrc.yaml` or `.yml`):
+
 ```yaml
 domain: http://localhost:11434/v1
 model: gemma2:9b-instruct
@@ -78,6 +85,7 @@ Use an environment variable for your API key. This is the safest approach and av
 - Also supported: `OPENAI_API_KEY`
 
 Examples:
+
 - bash/zsh:
   - `export OPENROUTER_API_KEY=sk-...` (place in `~/.bashrc` or `~/.zshrc` to persist)
 - fish:
@@ -88,6 +96,7 @@ Examples:
 CI/CD: store the key as a secret and inject it into the job environment.
 
 Auto‑init behavior
+
 - In interactive terminals, `openrouter ask|test|repl` will trigger `openrouter init` if no API key is available. Pass `--no-init` to skip.
 
 ## CLI Flags (Per‑Command)
@@ -96,12 +105,13 @@ Auto‑init behavior
   - `openrouter test --profile dev`
   - `openrouter ask --profile dev "Hello world"`
   - `openrouter repl --profile dev`
-  
+
 Note: To change the default domain or model, re‑run `openrouter init`. Per‑invocation model overrides have been removed to keep usage simple.
 
 ## Putting It Together: Precedence
 
 Example resolution for `openrouter ask --profile dev "Hi"` in a project with `.openrouterrc`:
+
 - CLI flags: profile=dev (wins over all below)
 - Env: API key from `OPENROUTER_API_KEY`, if set
 - Project `.openrouterrc`: overrides domain/model for this project
diff --git a/docs/CONTRIBUTING.md b/docs/CONTRIBUTING.md
index ae635fd..8c92b2d 100644
--- a/docs/CONTRIBUTING.md
+++ b/docs/CONTRIBUTING.md
@@ -3,10 +3,12 @@
 This project uses Conventional Commits and semantic‑release. Releases are automated from `main`; prereleases publish from feature branches.
 
 Basics
+
 - Write descriptive, minimal PRs that include a brief summary, usage examples, and updated docs when behavior changes.
 - Follow Conventional Commits (e.g., `feat: ...`, `fix: ...`, `docs: ...`, `refactor: ...`, `test: ...`). Use scopes when helpful: `feat(config): ...`.
 
 Local workflow
+
 - Install deps: `npm install`
 - Lint: `npm run lint`
 - Test: `npm test`
@@ -18,9 +20,11 @@ Local workflow
   - pre‑push: tests + build
 
 Tooling tips
+
 - If tests fail in a restricted shell due to worker/thread sandboxing, run them in CI or a normal local shell (where Vitest can spawn workers). CI is the source of truth.
 - If `npm pack --dry-run` fails due to cache permissions, use the local cache script: `npm run pack:dry` (sets `--cache ./.npm-cache`). To fix your user cache permanently, run: `sudo chown -R $(id -u):$(id -g) ~/.npm`.
 
 CI/Release
+
 - CI runs build, lint, and tests on Node 18.x and 20.x.
 - Release workflow uses semantic‑release with provenance enabled.
diff --git a/docs/DECISION.md b/docs/DECISION.md
index 042f880..55b387b 100644
--- a/docs/DECISION.md
+++ b/docs/DECISION.md
@@ -7,6 +7,7 @@ This document tracks major design choices, scope, and phased roadmap for the **o
 ## Iteration 1 (v0.1.x)
 
 ### Scope
+
 - npm-first TypeScript CLI (alias: `openrouter`).
 - Global config at `~/.config/openrouter-cli/config.json` (no project-local rc yet).
 - Defaults:
@@ -24,10 +25,12 @@ This document tracks major design choices, scope, and phased roadmap for the **o
   - One-shot `openrouter ask` command
 
 ### Security
+
 - Never log API keys.
 - Config file chmod 600 where possible.
 
 ### Out of Scope (queued for later)
+
 - Project-local `.openrouterrc` overrides.
 - Profiles (e.g. `--profile dev`).
 - Additional backends (Ollama, OpenAI, Anthropic) via `--domain` or `--backend`.
@@ -36,17 +39,20 @@ This document tracks major design choices, scope, and phased roadmap for the **o
 - Packaging expansion (Homebrew, winget, PyPI).
 
 ### Rationale
+
 - Optimize for frictionless `npm i -g` and `npx` demos for Node-first users.
 - Stay provider-agnostic via OpenAI-compatible API and configurable domain.
 - Ship streaming UX and exit semantics early for smooth demos.
 
 ### Next Checkpoints
+
 - Wire skeleton code into repo and basic CI (build + lint + test). ✅ Completed
 - Release automation: adopt semantic-release with Conventional Commits. ✅ Completed
 - Release from `main` via CI (no manual tagging). ✅ Completed (shipped as `v1.0.0`)
 - Plan Iteration 2: `.openrouterrc` & profiles. ✅ Completed
 
 ### Acceptance (Completed as v1.0.0)
+
 - All scope items delivered (CLI, config, defaults, `config`/`test`/`ask`/`repl` with streaming, API key handling).
 - Security constraints honored (no key logging; best‑effort chmod 600).
 - CI in place (build, lint, test) and production release automation configured.
@@ -57,6 +63,7 @@ This document tracks major design choices, scope, and phased roadmap for the **o
 ## Extension Roadmap
 
 ### Phase 1 — Tool Plugins (Manual Mode)
+
 - Local tool discovery:
   - `~/.config/openrouter-cli/tools.d/*.js`
   - `./.openrouter/tools/*.js`
@@ -77,6 +84,7 @@ This document tracks major design choices, scope, and phased roadmap for the **o
 - Deliverable: starter tool `web.search` (DuckDuckGo, no key).
 
 ### Phase 2 — Tool Calling (Auto Mode)
+
 - Expose tools as OpenAI-compatible `tools` (function-calling) in chat requests.
 - On `tool_call`, execute plugin and loop response as `tool` message.
 - Modes: `--tools off|manual|auto` (default manual).
@@ -84,6 +92,7 @@ This document tracks major design choices, scope, and phased roadmap for the **o
 - Evidence logging (opt-in) for compliance.
 
 ### Phase 3 — MCP Client Integration
+
 - Minimal MCP client (stdio first; WS/HTTP later).
 - Config:
   - `~/.config/openrouter-cli/mcp.d/*.json`
@@ -101,6 +110,7 @@ This document tracks major design choices, scope, and phased roadmap for the **o
 ---
 
 ## Cross-Cutting Later Items
+
 - `.openrouterrc` project overrides & profiles.
 - Keychain-backed secret storage (macOS/Win/gnome-keyring).
 - Enhanced packaging:
@@ -110,8 +120,8 @@ This document tracks major design choices, scope, and phased roadmap for the **o
 
 ---
 
-
 ### Release Process (v0.1.x)
+
 - Releasing
   - Stable releases occur on `main` via semantic-release. Conventional Commits decide bump.
   - Git tags (`vX.Y.Z`) and GitHub Releases are created by CI; npm publish is automated.
@@ -122,15 +132,16 @@ This document tracks major design choices, scope, and phased roadmap for the **o
   - commitlint + Conventional Commits enforced in PR and via local Husky hooks.
   - ESLint on staged files pre-commit; tests pre-push.
 
-
 ## Iteration 2 (v0.2.0) — Status
 
 ### Scope
+
 - Project-local `.openrouterrc` overrides (JSON and YAML).
 - Profiles (e.g. `--profile dev`) with per-profile overrides.
 - Init-driven defaults: provider/domain/model set via `openrouter init` (re-run to change).
 
 ### Design
+
 - Config precedence (effective):
   - CLI flags (e.g., `--profile`, output/stream options) > env vars (API key) > project `.openrouterrc` (JSON/YAML) > global config > built-in defaults.
 - Files:
@@ -141,21 +152,25 @@ This document tracks major design choices, scope, and phased roadmap for the **o
   - Global config can define `profiles` object; missing keys fall back to base config.
 
 ### CLI/UX
+
 - `--profile` available on `ask`, `repl`, and `test`.
 - `init` is the only way to change provider/domain/model (interactive wizard; auto-prompts in TTY when key is missing unless `--no-init`).
 - `config` is narrowed to listing and API key persistence (base or profile); domain/model/provider flags removed.
 - Removed per-invocation model override (`ask -m`).
 
 ### Security
+
 - Same key-handling guarantees; never log secrets.
 - Config file written with chmod 600 where possible; keys redacted in `--list` output.
 - Project file is not created automatically; only read if present.
 
 ### Tasks
+
 - Implement loader for `.openrouterrc` (JSON/YAML) with merge logic and precedence. ✅ Completed
 - Add profile resolver with fallback. ✅ Completed
 - Update help/docs and examples (consumer-focused README, CONFIGURATION). ✅ Completed
 - Add tests: URL join; YAML rc precedence; config redaction; streaming SSE parsing; ask error handling; CLI help shape. ✅ Completed
 
 ### Out of Scope for v0.2
+
 - TOML formats, keychain storage, remote profiles, tool plugins/MCP phases.
diff --git a/docs/TODO.md b/docs/TODO.md
index 33e9f86..071ba6e 100644
--- a/docs/TODO.md
+++ b/docs/TODO.md
@@ -1,6 +1,7 @@
 # Next TODOs (Iteration Planning)
 
 ## Top Priority — CLI Visual Polish
+
 - [ ] Discuss desired look/feel (Claude/Codex‑style): banner/logo, colors, spacing.
 - [ ] Pick libraries: `chalk` (colors), `boxen` (frames), `gradient-string` (optional), `ora` (spinners), `table`/`cli-table3` (tabular output).
 - [ ] Optional image/avatar in terminal (TTY only): evaluate `terminal-image` with fallback to ASCII art.
@@ -8,6 +9,7 @@
 - [ ] Accessibility: color‑safe palette, no-color fallback via `NO_COLOR`/TTY detection.
 
 ## Tests & Coverage
+
 - [ ] Add coverage thresholds in Vitest (start: 70% lines/branches; raise gradually).
 - [ ] Add tests:
   - [ ] ask (non‑stream) markdown snapshot (ANSI stripped before assert).
@@ -17,23 +19,28 @@
   - [ ] Optional: CLI E2E smoke (pack → run `--help`) in CI artifact.
 
 ## Error Handling & Timeouts
+
 - [ ] Add request timeout to fetch and surface timeouts clearly.
 - [ ] Normalize error output (network/DNS/401/429) with concise guidance.
 
 ## Init Flow Polish
+
 - [ ] Improve prompts copy; add final summary/confirm before save.
 - [ ] Consider non‑interactive flags for automation: `init --provider --domain --model --api-key --profile`.
 - [ ] Keep domain/model changes only via `init` (docs reflect this).
 
 ## Docs
+
 - [ ] README: add short, copy‑paste examples (ask md render, brief REPL transcript).
 - [ ] Consider a small GIF/screencast once UI polish lands.
 
 ## CI & Packaging
+
 - [ ] Upload coverage (e.g., Codecov) [optional].
 - [ ] Attach `npm pack --dry-run` tarball as CI artifact for review.
 
 ## Open Questions (to discuss)
+
 - [ ] Keep `.openrouterrc` project overrides long‑term or simplify later?
 - [ ] Image/avatar in terminal: ship ASCII art only, or inline images when supported?
 - [ ] Non‑interactive `init` flags: needed now for CI/scripts?
diff --git a/eslint.config.js b/eslint.config.js
index df0ac51..ede9fcc 100644
--- a/eslint.config.js
+++ b/eslint.config.js
@@ -1,59 +1,62 @@
 // ESLint flat config
-import js from "@eslint/js";
-import tseslint from "@typescript-eslint/eslint-plugin";
-import tsparser from "@typescript-eslint/parser";
-import globals from "globals";
+import js from '@eslint/js';
+import tseslint from '@typescript-eslint/eslint-plugin';
+import tsparser from '@typescript-eslint/parser';
+import globals from 'globals';
+import prettier from 'eslint-config-prettier';
 
 export default [
-  { ignores: ["dist/**", "node_modules/**"] },
+  { ignores: ['dist/**', 'node_modules/**'] },
   js.configs.recommended,
   {
-    files: ["**/*.js", "**/*.mjs", "**/*.cjs"],
+    files: ['**/*.js', '**/*.mjs', '**/*.cjs'],
     languageOptions: {
       ecmaVersion: 2022,
-      sourceType: "module",
+      sourceType: 'module',
       globals: {
         ...globals.node,
         ...globals.es2022,
       },
     },
     rules: {
-      "no-console": "off",
-      "no-undef": "off"
-    }
+      'no-console': 'off',
+      'no-undef': 'off',
+    },
   },
   {
-    files: ["**/*.ts"],
+    files: ['**/*.ts'],
     languageOptions: {
       ecmaVersion: 2022,
-      sourceType: "module",
+      sourceType: 'module',
       parser: tsparser,
       globals: {
         ...globals.node,
         ...globals.es2022,
-        fetch: "readonly",
-        Response: "readonly",
-        URL: "readonly",
-        ReadableStream: "readonly",
-        TextDecoder: "readonly"
+        fetch: 'readonly',
+        Response: 'readonly',
+        URL: 'readonly',
+        ReadableStream: 'readonly',
+        TextDecoder: 'readonly',
       },
     },
     plugins: {
-      "@typescript-eslint": tseslint,
+      '@typescript-eslint': tseslint,
     },
     rules: {
-      "no-console": "off",
-      "no-empty": ["error", { "allowEmptyCatch": true }],
+      'no-console': 'off',
+      'no-empty': ['error', { allowEmptyCatch: true }],
       // Use the TS-aware rule and disable the base one for TS files
-      "no-unused-vars": "off",
-      "@typescript-eslint/no-unused-vars": [
-        "error",
+      'no-unused-vars': 'off',
+      '@typescript-eslint/no-unused-vars': [
+        'error',
         {
-          "args": "after-used",
-          "argsIgnorePattern": "^_",
-          "varsIgnorePattern": "^_"
-        }
-      ]
-    }
-  }
+          args: 'after-used',
+          argsIgnorePattern: '^_',
+          varsIgnorePattern: '^_',
+        },
+      ],
+    },
+  },
+  // Disable formatting-related ESLint rules; let Prettier handle formatting
+  prettier,
 ];
diff --git a/existing_structure.md b/existing_structure.md
index d951726..3f88a39 100644
--- a/existing_structure.md
+++ b/existing_structure.md
@@ -63,6 +63,7 @@ src/index.ts ─▶ src/main.ts ─▶ commands
 - `safeJson(res)`: tries JSON, falls back to text.
 
 Notes:
+
 - Request-level timeouts and normalized error mapping are not yet implemented (planned in upcoming tasks).
 
 ## Rendering & UI
@@ -107,4 +108,3 @@ Notes:
 - Non-interactive flags for `init` (provider/domain/model/api-key/profile) as part of CI flows.
 - Documentation polish (README examples, REPL transcript, small screencast).
 - Optional CI packaging and coverage upload.
-
diff --git a/package-lock.json b/package-lock.json
index c0f6780..db88b91 100644
--- a/package-lock.json
+++ b/package-lock.json
@@ -39,9 +39,11 @@
         "@vitest/coverage-v8": "^2.1.1",
         "commitizen": "^4.3.0",
         "eslint": "^9.35.0",
+        "eslint-config-prettier": "^9.1.0",
         "globals": "^15.9.0",
         "husky": "^9.1.6",
         "lint-staged": "^15.2.10",
+        "prettier": "^3.3.3",
         "semantic-release": "^23.1.1",
         "ts-node": "^10.9.2",
         "typescript": "^5.9.2",
@@ -4810,6 +4812,19 @@
         }
       }
     },
+    "node_modules/eslint-config-prettier": {
+      "version": "9.1.2",
+      "resolved": "https://registry.npmjs.org/eslint-config-prettier/-/eslint-config-prettier-9.1.2.tgz",
+      "integrity": "sha512-iI1f+D2ViGn+uvv5HuHVUamg8ll4tN+JRHGc6IJi4TP9Kl976C57fzPXgseXNs8v0iA8aSJpHsTWjDb9QJamGQ==",
+      "dev": true,
+      "license": "MIT",
+      "bin": {
+        "eslint-config-prettier": "bin/cli.js"
+      },
+      "peerDependencies": {
+        "eslint": ">=7.0.0"
+      }
+    },
     "node_modules/eslint-scope": {
       "version": "8.4.0",
       "resolved": "https://registry.npmjs.org/eslint-scope/-/eslint-scope-8.4.0.tgz",
@@ -10757,6 +10772,22 @@
         "node": ">= 0.8.0"
       }
     },
+    "node_modules/prettier": {
+      "version": "3.6.2",
+      "resolved": "https://registry.npmjs.org/prettier/-/prettier-3.6.2.tgz",
+      "integrity": "sha512-I7AIg5boAr5R0FFtJ6rCfD+LFsWHp81dolrFD8S79U9tb8Az2nGrJncnMSnys+bpQJfRUzqs9hnA81OAA3hCuQ==",
+      "dev": true,
+      "license": "MIT",
+      "bin": {
+        "prettier": "bin/prettier.cjs"
+      },
+      "engines": {
+        "node": ">=14"
+      },
+      "funding": {
+        "url": "https://github.com/prettier/prettier?sponsor=1"
+      }
+    },
     "node_modules/pretty-ms": {
       "version": "9.2.0",
       "resolved": "https://registry.npmjs.org/pretty-ms/-/pretty-ms-9.2.0.tgz",
diff --git a/package.json b/package.json
index d207d6e..f3433e3 100644
--- a/package.json
+++ b/package.json
@@ -14,6 +14,8 @@
     "build": "tsc -p .",
     "dev": "node --loader ts-node/esm src/index.ts",
     "lint": "eslint \"src/**/*.{js,ts,mjs,cjs}\"",
+    "format": "prettier --write .",
+    "format:check": "prettier --check .",
     "test": "vitest run --reporter=default",
     "test:coverage": "vitest run --reporter=default --coverage",
     "test:watch": "vitest --watch",
@@ -77,16 +79,22 @@
     "@typescript-eslint/parser": "^8.42.0",
     "commitizen": "^4.3.0",
     "eslint": "^9.35.0",
+    "eslint-config-prettier": "^9.1.0",
     "globals": "^15.9.0",
     "husky": "^9.1.6",
     "lint-staged": "^15.2.10",
+    "prettier": "^3.3.3",
     "semantic-release": "^23.1.1",
     "ts-node": "^10.9.2",
     "typescript": "^5.9.2",
     "vitest": "^2.1.1"
   },
   "lint-staged": {
-    "src/**/*.{ts,js,mjs,cjs}": "eslint --fix --max-warnings=0"
+    "src/**/*.{ts,js,mjs,cjs}": [
+      "eslint --fix --max-warnings=0",
+      "prettier --write"
+    ],
+    "**/*.{json,md,yml,yaml}": "prettier --write"
   },
   "config": {
     "commitizen": {
diff --git a/scripts/prepack.mjs b/scripts/prepack.mjs
index d7e76ab..4d63467 100644
--- a/scripts/prepack.mjs
+++ b/scripts/prepack.mjs
@@ -12,4 +12,3 @@ async function main() {
 }
 
 main().catch(() => process.exit(0));
-
diff --git a/src/commands/models.ts b/src/commands/models.ts
index 5667971..47057db 100644
--- a/src/commands/models.ts
+++ b/src/commands/models.ts
@@ -2,7 +2,7 @@ import { Command } from 'commander';
 import { banner, palette } from '../shared/ui.js';
 import { resolveConfig } from '../shared/config.js';
 import { getDefaultConfig } from '../shared/env.js';
-import { fetchModelsCached, fuzzyIds} from '../shared/models.js';
+import { fetchModelsCached, fuzzyIds } from '../shared/models.js';
 import Table from 'cli-table3';
 
 export function registerModelsCommand(program: Command) {
@@ -16,7 +16,8 @@ export function registerModelsCommand(program: Command) {
       const domain = eff.domain || getDefaultConfig().domain;
       const apiKey = eff.apiKey; // do not print
 
-      const interactive = process.stdin.isTTY && process.stdout.isTTY && !opts.nonInteractive && !query;
+      // Default: interactive picker in TTY, even when a query is provided (query seeds initial results)
+      const interactive = process.stdin.isTTY && process.stdout.isTTY && !opts.nonInteractive;
       if (interactive) {
         try {
           console.log(banner());
@@ -28,7 +29,8 @@ export function registerModelsCommand(program: Command) {
           try {
             const list = await fetchModelsCached({ domain, apiKey });
             modelsList = list;
-            initial = list.slice(0, 25).map(m => m.id);
+            const base = query ? fuzzyIds(query, list, 25) : list.slice(0, 25).map((m) => m.id);
+            initial = base.length ? base : initial;
           } catch {}
           const stableSuggest = async (input: string) => {
             const q = input || '';
@@ -36,7 +38,8 @@ export function registerModelsCommand(program: Command) {
             const withTyped = q && !ids.includes(q) ? [q, ...ids] : ids;
             return toChoices(withTyped.length ? withTyped : initial);
           };
-          const toChoices = (ids: string[]) => ids.map(id => ({ name: id, value: id, message: id }));
+          const toChoices = (ids: string[]) =>
+            ids.map((id) => ({ name: id, value: id, message: id }));
           const ans = await promptFn({
             type: 'autocomplete',
             name: 'model',
@@ -62,7 +65,7 @@ export function registerModelsCommand(program: Command) {
         const ids = fuzzyIds(query || '', list, 25);
         const table = new Table({ head: ['ID', 'Name'] });
         for (const id of ids) {
-          const meta = list.find(m => m.id === id);
+          const meta = list.find((m) => m.id === id);
           table.push([id, meta?.name || '']);
         }
         console.log(table.toString());
diff --git a/src/index.ts b/src/index.ts
index 6e46497..b79d2a1 100644
--- a/src/index.ts
+++ b/src/index.ts
@@ -1,6 +1,6 @@
 #!/usr/bin/env node
-import { main } from "./main.js";
-import { logError } from "./shared/logger.js";
+import { main } from './main.js';
+import { logError } from './shared/logger.js';
 
 // Entrypoint
 main().catch(async (err) => {
diff --git a/src/main.ts b/src/main.ts
index 5abf275..e05e4a2 100644
--- a/src/main.ts
+++ b/src/main.ts
@@ -1,72 +1,132 @@
-import { Command } from "commander";
-import { readConfig, updateConfig, ensureConfigDir, updateProfile, resolveConfig } from "./shared/config.js";
-import type { CliConfig } from "./shared/config.js";
-import { getDefaultConfig, maskKey } from "./shared/env.js";
-import { testConnection, askOnce, ChatOptions, streamChat } from "./shared/openrouter.js";
-import { renderText, OutputFormat } from "./shared/format.js";
-import { startRepl } from "./repl.js";
-import { runInitWizard } from "./shared/init.js";
-import { attachStyledHelp, answerHeader, infoFooter, showSpinner } from "./shared/ui.js";
-import { registerModelsCommand } from "./commands/models.js";
+import { Command } from 'commander';
+import {
+  readConfig,
+  updateConfig,
+  ensureConfigDir,
+  updateProfile,
+  resolveConfig,
+  resetConfig,
+  overrideConfig,
+} from './shared/config.js';
+import type { CliConfig } from './shared/config.js';
+import { getDefaultConfig, maskKey } from './shared/env.js';
+import { testConnection, askOnce, ChatOptions, streamChat } from './shared/openrouter.js';
+import { renderText, OutputFormat } from './shared/format.js';
+import { startRepl } from './repl.js';
+import { runInitWizard } from './shared/init.js';
+import { attachStyledHelp, answerHeader, infoFooter, showSpinner } from './shared/ui.js';
+import { isPolicyError, handlePolicyError } from './shared/errors.js';
+import { registerModelsCommand } from './commands/models.js';
 
 export function buildProgram() {
   const program = new Command();
-  program
-    .name("openrouter")
-    .description("OpenRouter CLI")
-    .version("0.1.0");
+  program.name('openrouter').description('OpenRouter CLI').version('0.1.0');
+
+  // Attach styled help (banner + examples)
+  attachStyledHelp(program);
 
   // Attach styled help (banner + examples)
   attachStyledHelp(program);
 
   program
-    .command("config")
-    .description("Show configuration or update API key")
-    .option("--api-key <key>", "Persist API key (use env for ephemeral)")
-    .option("--profile <name>", "Select profile to read/update (default: base)")
-    .option("--list", "List profiles and current base config")
-    .action(async (opts: { apiKey?: string; profile?: string; list?: boolean }) => {
-      await ensureConfigDir();
-      if (opts.list) {
-        const cfg = await readConfig();
-        const redacted = JSON.parse(JSON.stringify(cfg));
-        if (redacted.apiKey) redacted.apiKey = maskKey(redacted.apiKey);
-        if (redacted.profiles) {
-          for (const p of Object.keys(redacted.profiles)) {
-            if (redacted.profiles[p]?.apiKey) redacted.profiles[p]!.apiKey = maskKey(redacted.profiles[p]!.apiKey as string);
+    .command('config')
+    .description('Show configuration or update API key')
+    .option('--api-key <key>', 'Persist API key (use env for ephemeral)')
+    .option('--profile <name>', 'Select profile to read/update (default: base)')
+    .option('--list', 'List profiles and current base config')
+    .option('--danger-reset', 'Danger: delete global config (debug)')
+    .option('--override-json <json>', 'Danger: replace global config with JSON (debug)')
+    .action(
+      async (opts: {
+        apiKey?: string;
+        profile?: string;
+        list?: boolean;
+        dangerReset?: boolean;
+        overrideJson?: string;
+      }) => {
+        await ensureConfigDir();
+        if (opts.dangerReset) {
+          await resetConfig();
+        }
+        if (opts.overrideJson) {
+          try {
+            const parsed = JSON.parse(opts.overrideJson) as CliConfig;
+            await overrideConfig(parsed);
+          } catch {
+            console.error('Invalid JSON for --override-json');
+            process.exitCode = 2;
+            return;
           }
         }
-        console.log(JSON.stringify(redacted, null, 2));
-        return;
-      }
+        if (opts.list) {
+          const cfg = await readConfig();
+          const redacted = JSON.parse(JSON.stringify(cfg));
+          if (redacted.apiKey) redacted.apiKey = maskKey(redacted.apiKey);
+          if (redacted.profiles) {
+            for (const p of Object.keys(redacted.profiles)) {
+              if (redacted.profiles[p]?.apiKey)
+                redacted.profiles[p]!.apiKey = maskKey(redacted.profiles[p]!.apiKey as string);
+            }
+          }
+          console.log(JSON.stringify(redacted, null, 2));
+          return;
+        }
 
-      const changes: Partial<CliConfig> = {};
-      // Only support updating API key here; domain/model are managed via `init`
-      if (opts.apiKey) changes.apiKey = opts.apiKey; // never log this
+        const changes: Partial<CliConfig> = {};
+        // Only support updating API key here; domain/model are managed via `init`
+        if (opts.apiKey) changes.apiKey = opts.apiKey; // never log this
 
-      if (Object.keys(changes).length > 0) {
-        if (opts.profile) {
-          await updateProfile(opts.profile, changes);
-        } else {
-          await updateConfig(changes);
+        if (Object.keys(changes).length > 0) {
+          if (opts.profile) {
+            await updateProfile(opts.profile, changes);
+          } else {
+            await updateConfig(changes);
+          }
         }
-      }
 
-      const cfg = await readConfig();
-      const redacted = { ...cfg, apiKey: cfg.apiKey ? maskKey(cfg.apiKey) : undefined };
-      console.log(JSON.stringify(redacted, null, 2));
-    });
+        const cfg = await readConfig();
+        const redacted = { ...cfg, apiKey: cfg.apiKey ? maskKey(cfg.apiKey) : undefined };
+        console.log(JSON.stringify(redacted, null, 2));
+
+        // If running in a TTY, drop into REPL with effective defaults
+        if (process.stdout.isTTY) {
+          try {
+            const eff = await resolveConfig(opts.profile);
+            const { ensureApiKey } = await import('./shared/auth.js');
+            const r = await ensureApiKey(opts.profile, true);
+            if (!('ok' in r) || !r.ok) {
+              console.error(r.message);
+              process.exitCode = 2;
+              return;
+            }
+            const apiKey = r.apiKey;
+            await startRepl({
+              apiKey,
+              domain: eff.domain ?? getDefaultConfig().domain,
+              initialModel: eff.model ?? getDefaultConfig().model,
+            });
+          } catch (error) {
+            console.error('Failed to start REPL:', error);
+            process.exitCode = 1;
+          }
+        }
+      }
+    );
 
   program
-    .command("test")
-    .description("Check API connectivity via /models")
-    .option("--profile <name>", "Use a named profile")
-    .option("--no-init", "Do not run interactive init when missing API key")
+    .command('test')
+    .description('Check API connectivity via /models')
+    .option('--profile <name>', 'Use a named profile')
+    .option('--no-init', 'Do not run interactive init when missing API key')
     .action(async (opts: { profile?: string; init?: boolean }) => {
       const cfg = await resolveConfig(opts.profile);
       const { ensureApiKey } = await import('./shared/auth.js');
       const r = await ensureApiKey(opts.profile, opts.init);
-      if (!('ok' in r) || !r.ok) { console.error(r.message); process.exitCode = 2; return; }
+      if (!('ok' in r) || !r.ok) {
+        console.error(r.message);
+        process.exitCode = 2;
+        return;
+      }
       const apiKey = r.apiKey;
       const domain = cfg.domain ?? getDefaultConfig().domain;
       const res = await testConnection({ domain, apiKey });
@@ -74,23 +134,37 @@ export function buildProgram() {
     });
 
   program
-    .command("ask")
+    .command('ask')
     .description("One-shot question. Use 'openrouter init' to change defaults.")
-    .argument("<prompt>", "User prompt")
-    .option("-s, --system <text>", "System prompt")
-    .option("--format <mode>", "Output format: auto|plain|md (default: auto)")
-    .option("--profile <name>", "Use a named profile")
-    .option("--no-stream", "Disable streaming output")
-    .option("--no-init", "Do not run interactive init when missing API key")
-    .action(async function (this: import('commander').Command, prompt: string, options: { system?: string; stream?: boolean; profile?: string; format?: OutputFormat; init?: boolean }) {
+    .argument('<prompt>', 'User prompt')
+    .option('-s, --system <text>', 'System prompt')
+    .option('--format <mode>', 'Output format: auto|plain|md (default: auto)')
+    .option('--profile <name>', 'Use a named profile')
+    .option('--no-stream', 'Disable streaming output')
+    .option('--no-init', 'Do not run interactive init when missing API key')
+    .action(async function (
+      this: import('commander').Command,
+      prompt: string,
+      options: {
+        system?: string;
+        stream?: boolean;
+        profile?: string;
+        format?: OutputFormat;
+        init?: boolean;
+      }
+    ) {
       const eff = await resolveConfig(options.profile);
       const { ensureApiKey } = await import('./shared/auth.js');
       const r = await ensureApiKey(options.profile, options.init);
-      if (!('ok' in r) || !r.ok) { console.error(r.message); process.exitCode = 2; return; }
+      if (!('ok' in r) || !r.ok) {
+        console.error(r.message);
+        process.exitCode = 2;
+        return;
+      }
       const apiKey = r.apiKey;
       const model = eff.model || getDefaultConfig().model;
       const domain = eff.domain || getDefaultConfig().domain;
-      const format: OutputFormat = (options.format as OutputFormat) || "md";
+      const format: OutputFormat = (options.format as OutputFormat) || 'md';
       // Default streaming OFF unless user explicitly passed --no-stream/--stream (we honor only explicit input for stream)
       const src = (this as any).getOptionValueSource?.('stream');
       const streamExplicit = src === 'cli' || src === 'env';
@@ -106,25 +180,68 @@ export function buildProgram() {
         let stopped = false;
         try {
           spinner.start();
-          await streamChat({
-            ...chatOptions,
-            onFirstToken: () => { if (!stopped) { spinner.stop(); stopped = true; } },
-            onDone: () => { if (!stopped) { spinner.stop(); stopped = true; } },
-          }, [{ role: "user", content: prompt }]);
-          process.stdout.write("\n");
+          await streamChat(
+            {
+              ...chatOptions,
+              onFirstToken: () => {
+                if (!stopped) {
+                  spinner.stop();
+                  stopped = true;
+                }
+              },
+              onDone: () => {
+                if (!stopped) {
+                  spinner.stop();
+                  stopped = true;
+                }
+              },
+            },
+            [{ role: 'user', content: prompt }]
+          );
+          process.stdout.write('\n');
+        } catch (err) {
+          if (isPolicyError(err)) {
+            await handlePolicyError({ where: 'ask-stream', tty: !!process.stdout.isTTY });
+            return;
+          }
+          throw err;
         } finally {
-          if (!stopped) { spinner.stop(); }
+          if (!stopped) {
+            spinner.stop();
+          }
         }
       } else {
         const spinner = showSpinner('Thinking…');
         try {
           spinner.start();
-          const out = await askOnce(chatOptions, [{ role: "user", content: prompt }]);
-          const pretty = renderText(out, { format, streaming: false });
+          const result = await askOnce(chatOptions, [{ role: 'user', content: prompt }]);
+          const pretty = renderText(result.text, { format, streaming: false });
           // Styled header and footer around non-stream result
-          process.stdout.write(answerHeader(model) + "\n");
-          process.stdout.write(pretty + "\n");
-          process.stdout.write(infoFooter({ model, domain }) + "\n");
+          process.stdout.write(answerHeader(model) + '\n');
+          process.stdout.write(pretty + '\n');
+
+          // Display usage information if available
+          if (result.usage) {
+            const usage = result.usage;
+            const tokens = usage.total_tokens || usage.completion_tokens + usage.prompt_tokens;
+            const cost = usage.total_cost || usage.cost;
+            let usageInfo = `\nTokens: ${tokens}`;
+            if (usage.prompt_tokens && usage.completion_tokens) {
+              usageInfo += ` (prompt: ${usage.prompt_tokens}, completion: ${usage.completion_tokens})`;
+            }
+            if (cost) {
+              usageInfo += ` | Cost: $${cost.toFixed(6)}`;
+            }
+            process.stdout.write(usageInfo + '\n');
+          }
+
+          process.stdout.write(infoFooter({ model, domain }) + '\n');
+        } catch (err) {
+          if (isPolicyError(err)) {
+            await handlePolicyError({ where: 'ask', tty: !!process.stdout.isTTY });
+            return;
+          }
+          throw err;
         } finally {
           spinner.stop();
         }
@@ -132,16 +249,22 @@ export function buildProgram() {
     });
 
   program
-    .command("repl")
-    .description("Interactive chat with streaming. Commands: exit, /model, /system, /format, /stream")
-    .option("-m, --model <name>", "Override model for this session")
-    .option("--profile <name>", "Use a named profile")
-    .option("--no-init", "Do not run interactive init when missing API key")
+    .command('repl')
+    .description(
+      'Interactive chat with streaming. Commands: exit, /model, /system, /format, /stream'
+    )
+    .option('-m, --model <name>', 'Override model for this session')
+    .option('--profile <name>', 'Use a named profile')
+    .option('--no-init', 'Do not run interactive init when missing API key')
     .action(async (options: { model?: string; profile?: string; init?: boolean }) => {
       const eff = await resolveConfig(options.profile);
       const { ensureApiKey } = await import('./shared/auth.js');
       const r = await ensureApiKey(options.profile, options.init);
-      if (!('ok' in r) || !r.ok) { console.error(r.message); process.exitCode = 2; return; }
+      if (!('ok' in r) || !r.ok) {
+        console.error(r.message);
+        process.exitCode = 2;
+        return;
+      }
       const apiKey = r.apiKey;
       await startRepl({
         apiKey,
@@ -151,10 +274,32 @@ export function buildProgram() {
     });
 
   program
-    .command("init")
-    .description("Interactive first-time setup (provider, domain, key, model, profile)")
+    .command('init')
+    .description('Interactive first-time setup (provider, domain, key, model, profile)')
     .action(async () => {
       await runInitWizard();
+      // After init, drop into REPL with effective defaults when in a TTY
+      if (process.stdout.isTTY) {
+        try {
+          const eff = await resolveConfig();
+          const { ensureApiKey } = await import('./shared/auth.js');
+          const r = await ensureApiKey(undefined, false); // avoid re-entering init from here
+          if (!('ok' in r) || !r.ok) {
+            console.error(r.message);
+            process.exitCode = 2;
+            return;
+          }
+          const apiKey = r.apiKey;
+          await startRepl({
+            apiKey,
+            domain: eff.domain ?? getDefaultConfig().domain,
+            initialModel: eff.model ?? getDefaultConfig().model,
+          });
+        } catch (error) {
+          console.error('Failed to start REPL after init:', error);
+          process.exitCode = 1;
+        }
+      }
     });
 
   // Additional commands
@@ -165,5 +310,10 @@ export function buildProgram() {
 
 export async function main() {
   const program = buildProgram();
+  if (process.argv.length <= 2) {
+    // No args: launch init flow by default
+    await program.parseAsync(['node', 'openrouter', 'init']);
+    return;
+  }
   await program.parseAsync(process.argv);
 }
diff --git a/src/repl.ts b/src/repl.ts
index da1b053..2e22ca3 100644
--- a/src/repl.ts
+++ b/src/repl.ts
@@ -1,9 +1,11 @@
-import readline from "node:readline";
-import { streamChat, askOnce } from "./shared/openrouter.js";
-import { renderText, OutputFormat } from "./shared/format.js";
-import { showSpinner } from "./shared/ui.js";
-import { logError } from "./shared/logger.js";
-import { styledPrompt, tipBox } from "./shared/ui.js";
+import readline from 'node:readline';
+import { streamChat, askOnce, getCredits } from './shared/openrouter.js';
+import { renderText, OutputFormat } from './shared/format.js';
+import { showSpinner } from './shared/ui.js';
+import { logError } from './shared/logger.js';
+import { styledPrompt, tipBox } from './shared/ui.js';
+// dynamic imports used when needed to avoid readline conflicts
+import { fetchModelsCached, fuzzyIds } from './shared/models.js';
 
 type ReplOptions = {
   apiKey: string;
@@ -14,108 +16,305 @@ type ReplOptions = {
 export async function startRepl(opts: ReplOptions) {
   let currentModel = opts.initialModel;
   let system: string | undefined;
-  let format: OutputFormat = "md"; // default rendered markdown for non-stream outputs
+  let format: OutputFormat = 'md'; // default rendered markdown for non-stream outputs
   let streaming = false;
-  const history: { role: "user" | "assistant"; content: string }[] = [];
+  const history: { role: 'user' | 'assistant'; content: string }[] = [];
+  let shouldExit = false;
 
-  const rl = readline.createInterface({ input: process.stdin, output: process.stdout, terminal: true, prompt: '', historySize: 100, escapeCodeTimeout: 50 });
-  const prompt = () => rl.setPrompt(styledPrompt(currentModel));
-  prompt();
-  rl.prompt();
+  // Session tracking
+  let sessionTokens = 0;
+  let sessionCost = 0;
+  let sessionRequests = 0;
 
-  console.log(tipBox());
+  async function promptUser(): Promise<void> {
+    if (shouldExit) return;
 
-  // Keep REPL alive on Ctrl+C (SIGINT); show prompt again
-  rl.on('SIGINT', () => {
-    process.stdout.write("\n");
-    rl.prompt();
-  });
+    const rl = readline.createInterface({
+      input: process.stdin,
+      output: process.stdout,
+      terminal: true,
+      prompt: '',
+      historySize: 100,
+      escapeCodeTimeout: 50,
+      removeHistoryDuplicates: true,
+    });
 
-  rl.on("line", async (line) => {
-    const input = line.trim();
-    if (!input) {
-      rl.prompt();
-      return;
-    }
-    if (input === "exit") {
-      rl.close();
-      return;
-    }
-    if (input.startsWith("/model ")) {
-      currentModel = input.slice("/model ".length).trim();
-      prompt();
-      rl.prompt();
-      return;
-    }
-    if (input.startsWith("/system ")) {
-      system = input.slice("/system ".length).trim();
-      console.log("[system set]");
-      rl.prompt();
-      return;
+    const prompt = () => rl.setPrompt(styledPrompt(currentModel));
+    prompt();
+
+    if (history.length === 0) {
+      console.log(tipBox());
     }
-    if (input.startsWith("/format ")) {
-      const val = input.slice("/format ".length).trim();
-      if (val === "md" || val === "plain") {
-        format = val;
-        console.log(`[format: ${format}]`);
-      } else {
-        console.log("Usage: /format md|plain");
+
+    rl.on('line', async (line) => {
+      const input = line.trim();
+
+      if (!input) {
+        rl.prompt();
+        return;
+      }
+      if (input === 'exit') {
+        shouldExit = true;
+        rl.close();
+        return;
       }
-      rl.prompt();
-      return;
-    }
-    if (input.startsWith("/stream ")) {
-      const val = input.slice("/stream ".length).trim();
-      if (val === "on") streaming = true;
-      else if (val === "off") streaming = false;
-      else console.log("Usage: /stream on|off");
-      console.log(`[stream: ${streaming ? "on" : "off"}]`);
-      rl.prompt();
-      return;
-    }
 
-    const userMsg = { role: "user" as const, content: input };
-    history.push(userMsg);
-    try {
-      if (streaming) {
-        const spinner = showSpinner('Thinking');
-        let stopped = false;
+      // Close the readline interface after getting input
+      rl.close();
+      if (input === '/model') {
         try {
+          const spinner = showSpinner('Loading models…');
           spinner.start();
-          await streamChat({
-            domain: opts.domain,
-            apiKey: opts.apiKey,
-            model: currentModel,
-            system,
-            stream: true,
-            onFirstToken: () => { if (!stopped) { spinner.stop(); stopped = true; } },
-            onDone: () => { if (!stopped) { spinner.stop(); stopped = true; } },
-          }, [
-            ...history,
-          ]);
-          process.stdout.write("\n");
-        } finally {
-          if (!stopped) { spinner.stop(); }
+          const list = await fetchModelsCached({ domain: opts.domain, apiKey: opts.apiKey });
+          spinner.stop();
+
+          // Create a new readline interface for model selection
+          const modelRl = readline.createInterface({
+            input: process.stdin,
+            output: process.stdout,
+            terminal: true,
+          });
+
+          const askModel = (q: string) =>
+            new Promise<string>((res) => modelRl.question(q, (ans) => res(ans.trim())));
+
+          const query = await askModel('Search models (>=2 chars, blank to cancel): ');
+          if (!query) {
+            modelRl.close();
+            await promptUser();
+            return;
+          }
+
+          const ids = fuzzyIds(query, list, 10);
+          if (!ids.length) {
+            console.log('No matches.');
+            modelRl.close();
+            await promptUser();
+            return;
+          }
+
+          console.log('Matches:');
+          ids.forEach((id, i) => {
+            const meta = list.find((m) => m.id === id);
+            const name = meta?.name ? ` — ${meta.name}` : '';
+            console.log(`${i + 1}. ${id}${name}`);
+          });
+
+          const sel = await askModel(`Pick 1-${ids.length} or type a model id: `);
+          let chosen = '';
+          const n = Number(sel);
+          if (sel && Number.isInteger(n) && n >= 1 && n <= ids.length) {
+            chosen = ids[n - 1];
+          } else if (sel) {
+            chosen = sel;
+          }
+
+          if (chosen) {
+            currentModel = chosen;
+            console.log(`[model: ${currentModel}]`);
+          }
+
+          modelRl.close();
+        } catch (e) {
+          await logError(e, 'repl-model-picker');
+          console.log("Tip: run 'openrouter models' in another terminal to browse models.");
         }
-      } else {
-        const spinner = showSpinner('Thinking…');
+
+        // Ensure clean terminal state before restarting
+        process.stdout.write('\n');
+        await promptUser();
+        return;
+      }
+      if (input.startsWith('/model ')) {
+        currentModel = input.slice('/model '.length).trim();
+        console.log(`[model: ${currentModel}]`);
+        await promptUser();
+        return;
+      }
+      if (input.startsWith('/system ')) {
+        system = input.slice('/system '.length).trim();
+        console.log('[system set]');
+        await promptUser();
+        return;
+      }
+      if (input.startsWith('/format ')) {
+        const val = input.slice('/format '.length).trim();
+        if (val === 'md' || val === 'plain') {
+          format = val;
+          console.log(`[format: ${format}]`);
+        } else {
+          console.log('Usage: /format md|plain');
+        }
+        await promptUser();
+        return;
+      }
+      if (input.startsWith('/stream ')) {
+        const val = input.slice('/stream '.length).trim();
+        if (val === 'on') streaming = true;
+        else if (val === 'off') streaming = false;
+        else console.log('Usage: /stream on|off');
+        console.log(`[stream: ${streaming ? 'on' : 'off'}]`);
+        await promptUser();
+        return;
+      }
+      if (input === '/stats') {
+        console.log(`\nSession Statistics:`);
+        console.log(`Model: ${currentModel}`);
+        console.log(`Total tokens: ${sessionTokens}`);
+        console.log(`Total requests: ${sessionRequests}`);
+        if (sessionCost > 0) {
+          console.log(`Total cost: $${sessionCost.toFixed(6)}`);
+        } else {
+          console.log(`Total cost: Free`);
+        }
+        console.log(
+          `Average tokens per request: ${sessionRequests > 0 ? Math.round(sessionTokens / sessionRequests) : 0}`
+        );
+        await promptUser();
+        return;
+      }
+      if (input === '/billing') {
         try {
+          const spinner = showSpinner('Fetching billing info…');
           spinner.start();
-          const text = await askOnce({ domain: opts.domain, apiKey: opts.apiKey, model: currentModel, system, stream: false }, [
-            ...history,
-          ]);
-          const pretty = renderText(text, { format, streaming: false });
-          process.stdout.write(pretty + "\n");
-        } finally {
+          const credits = await getCredits({ domain: opts.domain, apiKey: opts.apiKey });
           spinner.stop();
+
+          console.log(`\nOpenRouter Account Credits:`);
+          if (credits.data) {
+            const data = credits.data;
+            console.log(`Total credits purchased: $${(data.total_credits || 0).toFixed(6)}`);
+            console.log(`Total usage: $${(data.total_usage || 0).toFixed(6)}`);
+            console.log(
+              `Current balance: $${((data.total_credits || 0) - (data.total_usage || 0)).toFixed(6)}`
+            );
+            if (data.credit_limit) {
+              console.log(`Credit limit: $${data.credit_limit.toFixed(6)}`);
+            }
+          } else {
+            console.log(`Credits info: ${JSON.stringify(credits, null, 2)}`);
+          }
+        } catch (err) {
+          console.error('Failed to fetch billing info:', err);
         }
+        await promptUser();
+        return;
       }
-    } catch (err) {
-      await logError(err, 'repl');
-      console.error('A technical issue occurred. Please try again.');
-    }
+
+      const userMsg = { role: 'user' as const, content: input };
+      history.push(userMsg);
+
+      // Process the message and then prompt
+      try {
+        if (streaming) {
+          const spinner = showSpinner('Thinking');
+          let stopped = false;
+          try {
+            spinner.start();
+            await streamChat(
+              {
+                domain: opts.domain,
+                apiKey: opts.apiKey,
+                model: currentModel,
+                system,
+                stream: true,
+                onFirstToken: () => {
+                  if (!stopped) {
+                    spinner.stop();
+                    stopped = true;
+                  }
+                },
+                onDone: () => {
+                  if (!stopped) {
+                    spinner.stop();
+                    stopped = true;
+                  }
+                },
+              },
+              [...history]
+            );
+            process.stdout.write('\n');
+          } finally {
+            if (!stopped) {
+              spinner.stop();
+            }
+          }
+        } else {
+          const spinner = showSpinner('Thinking…');
+          try {
+            spinner.start();
+            const result = await askOnce(
+              {
+                domain: opts.domain,
+                apiKey: opts.apiKey,
+                model: currentModel,
+                system,
+                stream: false,
+              },
+              [...history]
+            );
+            const pretty = renderText(result.text, { format, streaming: false });
+            process.stdout.write(pretty + '\n');
+
+            // Display usage information if available
+            if (result.usage) {
+              const usage = result.usage;
+              const tokens = usage.total_tokens || usage.completion_tokens + usage.prompt_tokens;
+              const cost = usage.total_cost || usage.cost || 0;
+
+              // Update session totals
+              sessionTokens += tokens;
+              sessionCost += cost;
+              sessionRequests += 1;
+
+              let usageInfo = `\nTokens: ${tokens}`;
+              if (usage.prompt_tokens && usage.completion_tokens) {
+                usageInfo += ` (prompt: ${usage.prompt_tokens}, completion: ${usage.completion_tokens})`;
+              }
+
+              if (cost > 0) {
+                usageInfo += ` | Cost: $${cost.toFixed(6)}`;
+              } else {
+                usageInfo += ` | Free`;
+              }
+
+              // Show session totals
+              usageInfo += `\nSession: ${sessionTokens} tokens`;
+              if (sessionCost > 0) {
+                usageInfo += `, $${sessionCost.toFixed(6)} total`;
+              } else {
+                usageInfo += `, free`;
+              }
+              usageInfo += ` (${sessionRequests} requests)`;
+
+              process.stdout.write(usageInfo + '\n');
+            }
+          } finally {
+            spinner.stop();
+          }
+        }
+      } catch (err) {
+        if ((await import('./shared/errors.js')).isPolicyError(err)) {
+          const { handlePolicyError } = await import('./shared/errors.js');
+          await handlePolicyError({
+            where: 'repl',
+            tty: !!process.stdout.isTTY,
+            interactivePrompt: false,
+          });
+        } else {
+          await logError(err, 'repl');
+          console.error('A technical issue occurred. Please try again.');
+        }
+      }
+
+      // Continue with next prompt
+      await promptUser();
+    });
+
+    // Start the initial prompt
     rl.prompt();
-  });
+    await new Promise<void>((resolve) => rl.once('close', resolve));
+  }
 
-  await new Promise<void>((resolve) => rl.on("close", () => resolve()));
+  await promptUser();
 }
diff --git a/src/shared/auth.ts b/src/shared/auth.ts
index 6fe3a93..55b2147 100644
--- a/src/shared/auth.ts
+++ b/src/shared/auth.ts
@@ -4,9 +4,13 @@ import { runInitWizard } from './init.js';
 
 export type EnsureApiKeyResult = { ok: true; apiKey: string } | { ok: false; message: string };
 
-const MISSING_KEY_MSG = "Missing API key. Set OPENROUTER_API_KEY / OPENAI_API_KEY or run 'openrouter init'.";
+const MISSING_KEY_MSG =
+  "Missing API key. Set OPENROUTER_API_KEY / OPENAI_API_KEY or run 'openrouter init'.";
 
-export async function ensureApiKey(profile?: string, allowInit?: boolean): Promise<EnsureApiKeyResult> {
+export async function ensureApiKey(
+  profile?: string,
+  allowInit?: boolean
+): Promise<EnsureApiKeyResult> {
   const eff = await resolveConfig(profile);
   let apiKey = getApiKey(await readConfig()) || eff.apiKey;
 
@@ -21,4 +25,3 @@ export async function ensureApiKey(profile?: string, allowInit?: boolean): Promi
   if (!apiKey) return { ok: false, message: MISSING_KEY_MSG };
   return { ok: true, apiKey };
 }
-
diff --git a/src/shared/config.ts b/src/shared/config.ts
index 063ea98..36ade32 100644
--- a/src/shared/config.ts
+++ b/src/shared/config.ts
@@ -1,6 +1,6 @@
-import fs from "node:fs/promises";
-import path from "node:path";
-import os from "node:os";
+import fs from 'node:fs/promises';
+import path from 'node:path';
+import os from 'node:os';
 
 export type CliConfig = {
   domain?: string;
@@ -10,8 +10,8 @@ export type CliConfig = {
   profiles?: Record<string, Omit<CliConfig, 'profiles'>>;
 };
 
-const CONFIG_DIR = path.join(os.homedir(), ".config", "openrouter-cli");
-const CONFIG_FILE = path.join(CONFIG_DIR, "config.json");
+const CONFIG_DIR = path.join(os.homedir(), '.config', 'openrouter-cli');
+const CONFIG_FILE = path.join(CONFIG_DIR, 'config.json');
 
 export async function ensureConfigDir() {
   await fs.mkdir(CONFIG_DIR, { recursive: true });
@@ -23,11 +23,11 @@ export async function ensureConfigDir() {
 
 export async function readConfig(): Promise<CliConfig> {
   try {
-    const txt = await fs.readFile(CONFIG_FILE, "utf8");
+    const txt = await fs.readFile(CONFIG_FILE, 'utf8');
     const json = JSON.parse(txt) as CliConfig;
     return json;
   } catch (err: any) {
-    if (err?.code === "ENOENT") return {};
+    if (err?.code === 'ENOENT') return {};
     throw err;
   }
 }
@@ -50,7 +50,6 @@ export async function updateProfile(profile: string, patch: Partial<CliConfig>)
   const current = await readConfig();
   const profiles = { ...(current.profiles || {}) } as NonNullable<CliConfig['profiles']>;
   const cur = profiles[profile] || {};
-  
   const { profiles: _, ...safePatch } = patch;
   profiles[profile] = { ...cur, ...safePatch };
   await updateConfig({ profiles });
@@ -59,18 +58,18 @@ export async function updateProfile(profile: string, patch: Partial<CliConfig>)
 // Project-local overrides: .openrouterrc(.json|.yaml|.yml)
 export async function readProjectRc(cwd = process.cwd()): Promise<Partial<CliConfig>> {
   const candidates = [
-    path.join(cwd, ".openrouterrc"),
-    path.join(cwd, ".openrouterrc.json"),
-    path.join(cwd, ".openrouterrc.yaml"),
-    path.join(cwd, ".openrouterrc.yml"),
+    path.join(cwd, '.openrouterrc'),
+    path.join(cwd, '.openrouterrc.json'),
+    path.join(cwd, '.openrouterrc.yaml'),
+    path.join(cwd, '.openrouterrc.yml'),
   ];
   for (const file of candidates) {
     try {
-      const txt = await fs.readFile(file, "utf8");
+      const txt = await fs.readFile(file, 'utf8');
       const parsed = await parseRc(txt, path.extname(file));
       return (parsed || {}) as Partial<CliConfig>;
     } catch (err: any) {
-      if (err?.code === "ENOENT") continue;
+      if (err?.code === 'ENOENT') continue;
       throw err;
     }
   }
@@ -79,8 +78,8 @@ export async function readProjectRc(cwd = process.cwd()): Promise<Partial<CliCon
 
 async function parseRc(text: string, ext: string) {
   try {
-    if (ext === ".yaml" || ext === ".yml") {
-      const { load } = await import("js-yaml");
+    if (ext === '.yaml' || ext === '.yml') {
+      const { load } = await import('js-yaml');
       return load(text) as unknown;
     }
     // Try JSON first for no-extension or .json
@@ -89,7 +88,7 @@ async function parseRc(text: string, ext: string) {
     // Fallback: if no extension, try YAML
     if (!ext) {
       try {
-        const { load } = await import("js-yaml");
+        const { load } = await import('js-yaml');
         return load(text) as unknown;
       } catch {}
     }
@@ -111,3 +110,23 @@ export async function resolveConfig(profile?: string): Promise<ResolvedConfig> {
   const prof = profile ? global.profiles?.[profile] || {} : {};
   return { ...base, ...prof, ...project };
 }
+
+// Danger helpers for debugging/testing
+export async function resetConfig(): Promise<boolean> {
+  try {
+    await fs.unlink(paths.CONFIG_FILE);
+    return true;
+  } catch (e: any) {
+    if (e?.code === 'ENOENT') return false;
+    throw e;
+  }
+}
+
+export async function overrideConfig(newConfig: CliConfig): Promise<void> {
+  await ensureConfigDir();
+  const data = JSON.stringify(newConfig, null, 2);
+  await fs.writeFile(paths.CONFIG_FILE, data, { mode: 0o600 });
+  try {
+    await fs.chmod(paths.CONFIG_FILE, 0o600);
+  } catch {}
+}
diff --git a/src/shared/env.ts b/src/shared/env.ts
index 04e74bc..0dc093a 100644
--- a/src/shared/env.ts
+++ b/src/shared/env.ts
@@ -1,9 +1,9 @@
-import type { CliConfig } from "./config.js";
+import type { CliConfig } from './config.js';
 
 export function getDefaultConfig() {
   return {
-    domain: "https://openrouter.ai/api/v1",
-    model: "meta-llama/llama-3.1-8b-instruct",
+    domain: 'https://openrouter.ai/api/v1',
+    model: 'meta-llama/llama-3.1-8b-instruct',
   } as const;
 }
 
@@ -14,7 +14,6 @@ export function getApiKey(cfg: CliConfig): string | undefined {
 
 export function maskKey(key: string) {
   if (!key) return key;
-  if (key.length <= 8) return "*".repeat(key.length);
-  return key.slice(0, 4) + "****" + key.slice(-4);
+  if (key.length <= 8) return '*'.repeat(key.length);
+  return key.slice(0, 4) + '****' + key.slice(-4);
 }
-
diff --git a/src/shared/errors.ts b/src/shared/errors.ts
new file mode 100644
index 0000000..3c81cfe
--- /dev/null
+++ b/src/shared/errors.ts
@@ -0,0 +1,60 @@
+import { logError } from './logger.js';
+import { spawn } from 'node:child_process';
+
+export const PRIVACY_URL = 'https://openrouter.ai/settings/privacy';
+
+export function isPolicyError(err: unknown): boolean {
+  const msg = err instanceof Error ? err.message : String(err);
+  return /data policy/i.test(msg) || /Free model publication/i.test(msg);
+}
+
+export async function handlePolicyError(opts: {
+  where: string;
+  tty: boolean;
+  interactivePrompt?: boolean;
+}): Promise<void> {
+  const hint = `This model may require enabling free endpoints that can publish prompts.\nOpen privacy settings: ${PRIVACY_URL}`;
+  // Always log details; keep terminal output friendly
+  await logError(new Error(`[policy] in ${opts.where}`));
+  if (!opts.tty) {
+    console.error(hint + "\nTip: run 'openrouter models' to pick another model.");
+    return;
+  }
+  if (opts.interactivePrompt === false) {
+    console.error(hint + "\nTip: run 'openrouter models' to pick another model.");
+    return;
+  }
+  try {
+    const enq = await import('enquirer');
+    const promptFn: any = (enq as any).prompt ?? (enq as any).default?.prompt;
+    if (typeof promptFn === 'function') {
+      console.error(hint);
+      const ans = await promptFn({
+        type: 'confirm',
+        name: 'open',
+        message: 'Open settings now?',
+        initial: true,
+      });
+      if (ans?.open) await openUrl(PRIVACY_URL);
+      else console.error("You can also run 'openrouter models' to choose a different model.");
+      return;
+    }
+  } catch {
+    // ignore
+  }
+  console.error(hint);
+}
+
+async function openUrl(url: string) {
+  const platform = process.platform;
+  const trySpawn = (cmd: string, args: string[]) =>
+    new Promise<void>((resolve) => {
+      const p = spawn(cmd, args, { stdio: 'ignore', detached: true });
+      p.on('error', () => resolve());
+      p.unref();
+      resolve();
+    });
+  if (platform === 'darwin') return trySpawn('open', [url]);
+  if (platform === 'win32') return trySpawn('cmd', ['/c', 'start', '', url]);
+  return trySpawn('xdg-open', [url]);
+}
diff --git a/src/shared/format.ts b/src/shared/format.ts
index d9c08c7..299d78e 100644
--- a/src/shared/format.ts
+++ b/src/shared/format.ts
@@ -1,11 +1,14 @@
-export type OutputFormat = "auto" | "plain" | "md";
+export type OutputFormat = 'auto' | 'plain' | 'md';
 
 // Very small markdown → ANSI formatter (headings, lists, code)
-export function renderText(input: string, opts: { format: OutputFormat; streaming: boolean }): string {
+export function renderText(
+  input: string,
+  opts: { format: OutputFormat; streaming: boolean }
+): string {
   const mode = opts.format;
   if (opts.streaming) return input; // keep streaming plain for responsiveness
-  if (mode === "plain") return input;
-  if (mode === "md" || mode === "auto") return toAnsiMarkdown(input);
+  if (mode === 'plain') return input;
+  if (mode === 'md' || mode === 'auto') return toAnsiMarkdown(input);
   return input;
 }
 
@@ -17,29 +20,29 @@ export function toAnsiMarkdown(md: string): string {
     let line = lines[i];
     if (/^```/.test(line.trim())) {
       inFence = !inFence;
-      out.push(ansiDim(""));
+      out.push(ansiDim(''));
       continue;
     }
     if (inFence) {
-      out.push(ansiCyan("  " + line));
+      out.push(ansiCyan('  ' + line));
       continue;
     }
     // Headings
     if (/^###\s+/.test(line)) {
-      out.push(ansiBold(line.replace(/^###\s+/, "").trim()));
+      out.push(ansiBold(line.replace(/^###\s+/, '').trim()));
       continue;
     }
     if (/^##\s+/.test(line)) {
-      out.push(ansiBold(line.replace(/^##\s+/, "").trim()));
+      out.push(ansiBold(line.replace(/^##\s+/, '').trim()));
       continue;
     }
     if (/^#\s+/.test(line)) {
-      out.push(ansiBold(line.replace(/^#\s+/, "").trim()));
+      out.push(ansiBold(line.replace(/^#\s+/, '').trim()));
       continue;
     }
     // Lists
     if (/^\s*[-*+]\s+/.test(line)) {
-      line = line.replace(/^\s*[-*+]\s+/, "  • ");
+      line = line.replace(/^\s*[-*+]\s+/, '  • ');
       // fall through for inline formatting below
     }
     // Inline code: `code`
@@ -48,23 +51,29 @@ export function toAnsiMarkdown(md: string): string {
     line = line.replace(/\*\*([^*]+)\*\*/g, (_, m1: string) => ansiBold(m1));
     line = line.replace(/__([^_]+)__/g, (_, m1: string) => ansiBold(m1));
     // Inline italic: *text* or _text_ (basic, non-greedy, ignore escaped)
-    line = line.replace(/(^|[^\\])\*([^*\s][^*]*?)\*(?!\*)/g, (_m, p1: string, m1: string) => p1 + ansiItalic(m1));
-    line = line.replace(/(^|[^\\])_([^_\s][^_]*)_(?!_)/g, (_m, p1: string, m1: string) => p1 + ansiItalic(m1));
+    line = line.replace(
+      /(^|[^\\])\*([^*\s][^*]*?)\*(?!\*)/g,
+      (_m, p1: string, m1: string) => p1 + ansiItalic(m1)
+    );
+    line = line.replace(
+      /(^|[^\\])_([^_\s][^_]*)_(?!_)/g,
+      (_m, p1: string, m1: string) => p1 + ansiItalic(m1)
+    );
     out.push(line);
   }
-  return out.join("\n");
+  return out.join('\n');
 }
 
 // Minimal ANSI helpers
 function ansiBold(s: string) {
-  return "\x1b[1m" + s + "\x1b[22m";
+  return '\x1b[1m' + s + '\x1b[22m';
 }
 function ansiDim(s: string) {
-  return "\x1b[2m" + s + "\x1b[22m";
+  return '\x1b[2m' + s + '\x1b[22m';
 }
 function ansiCyan(s: string) {
-  return "\x1b[36m" + s + "\x1b[39m";
+  return '\x1b[36m' + s + '\x1b[39m';
 }
 function ansiItalic(s: string) {
-  return "\x1b[3m" + s + "\x1b[23m";
+  return '\x1b[3m' + s + '\x1b[23m';
 }
diff --git a/src/shared/init.ts b/src/shared/init.ts
index 5e58da3..08d1b69 100644
--- a/src/shared/init.ts
+++ b/src/shared/init.ts
@@ -1,15 +1,20 @@
 import readline from 'node:readline';
-import { updateConfig, updateProfile } from './config.js';
+import { updateConfig, updateProfile, readConfig } from './config.js';
 import type { CliConfig } from './config.js';
-import { getDefaultConfig } from './env.js';
+import { getDefaultConfig, getApiKey } from './env.js';
 import { testConnection } from './openrouter.js';
 import { fetchModelsCached, fuzzyIds } from './models.js';
-import { banner } from './ui.js';
+import { banner, warnBox } from './ui.js';
 
 type Provider = 'openrouter' | 'openai' | 'custom';
 
 function choosePreset(p: Provider) {
-  if (p === 'openai') return { provider: 'openai' as const, domain: 'https://api.openai.com/v1', model: 'gpt-4o-mini' };
+  if (p === 'openai')
+    return {
+      provider: 'openai' as const,
+      domain: 'https://api.openai.com/v1',
+      model: 'gpt-4o-mini',
+    };
   if (p === 'custom') return { provider: 'custom' as const, domain: '', model: '' };
   // default: openrouter
   const d = getDefaultConfig();
@@ -17,7 +22,11 @@ function choosePreset(p: Provider) {
 }
 
 export async function runInitWizard(): Promise<boolean> {
-  const rl = readline.createInterface({ input: process.stdin, output: process.stdout, terminal: true });
+  const rl = readline.createInterface({
+    input: process.stdin,
+    output: process.stdout,
+    terminal: true,
+  });
   const ask = (q: string) => new Promise<string>((res) => rl.question(q, (ans) => res(ans.trim())));
   const askHidden = (q: string) => ask(q); // fallback: no masking in this environment
 
@@ -25,14 +34,20 @@ export async function runInitWizard(): Promise<boolean> {
   const provider: Provider = 'openrouter';
   const preset = choosePreset(provider);
 
-  const domain = (await ask(`API domain (default: ${preset.domain || 'none'}): `)) || preset.domain;
-  let model = (await ask(`Default model (default: ${preset.model || 'none'}): `)) || preset.model;
-
-  let apiKey = process.env.OPENROUTER_API_KEY || process.env.OPENAI_API_KEY || '';
-  if (!apiKey) {
-    apiKey = await askHidden('API key (leave blank to skip): ');
+  // For MCP we only support OpenRouter, so use the default domain without prompting
+  const domain = preset.domain;
+  if (process.stdout.isTTY) {
+    console.log(`Using API domain: ${domain}`);
+  }
+  // Prefer interactive fuzzy search for model in TTY; fall back to plain prompt in non-TTY
+  let model = preset.model;
+  if (!process.stdout.isTTY) {
+    model = (await ask(`Default model (default: ${preset.model || 'none'}): `)) || preset.model;
   }
 
+  let apiKey = getApiKey(await readConfig()) || '';
+  if (!apiKey) apiKey = await askHidden('API key (leave blank to skip): ');
+
   const profile = await ask('Profile name (optional): ');
 
   // Interactive model picker (TTY only)
@@ -47,15 +62,15 @@ export async function runInitWizard(): Promise<boolean> {
       try {
         const list = await fetchModelsCached({ domain, apiKey: apiKey || undefined });
         modelsList = list;
-        initial = list.slice(0, 25).map(m => m.id);
+        initial = list.slice(0, 25).map((m) => m.id);
       } catch {}
       const stableSuggest = async (input: string) => {
         const q = input || '';
-        const ids = modelsList.length ? fuzzyIds(q, modelsList as any) : [] as string[];
+        const ids = modelsList.length ? fuzzyIds(q, modelsList as any) : ([] as string[]);
         const withTyped = q && !ids.includes(q) ? [q, ...ids] : ids;
         return toChoices(withTyped.length ? withTyped : initial);
       };
-      const toChoices = (ids: string[]) => ids.map(id => ({ name: id, value: id, message: id }));
+      const toChoices = (ids: string[]) => ids.map((id) => ({ name: id, value: id, message: id }));
       const ans = await promptFn({
         type: 'autocomplete',
         name: 'model',
@@ -85,16 +100,30 @@ export async function runInitWizard(): Promise<boolean> {
     try {
       await testConnection({ domain, apiKey });
       console.log('✓ Connection OK');
-    } catch (e) {
-      console.log('! Connection failed:', e instanceof Error ? e.message : String(e));
+    } catch {
+      console.log(warnBox('Connection failed. Please verify your API key and settings.'));
+      console.log('Visit https://openrouter.ai/keys to create a key.');
+      console.log(
+        "Set via env: export OPENROUTER_API_KEY='sk-...' or persist with: openrouter config --api-key sk-..."
+      );
       const ans = (await ask('Save settings anyway? [y/N]: ')).toLowerCase();
       save = ans === 'y' || ans === 'yes';
     }
+  } else if (!apiKey) {
+    console.log(warnBox('An API key is required to connect.'));
+    console.log('Create a key at https://openrouter.ai/keys');
+    console.log(
+      "Set via env: export OPENROUTER_API_KEY='sk-...' or persist with: openrouter config --api-key sk-..."
+    );
   }
 
   if (save) {
     // Persist selections. If a profile is provided, write under that profile; otherwise write to base config.
-    const changes: Partial<CliConfig> = { provider, domain: domain || undefined, model: model || undefined };
+    const changes: Partial<CliConfig> = {
+      provider,
+      domain: domain || undefined,
+      model: model || undefined,
+    };
     if (apiKey) changes.apiKey = apiKey;
     if (profile) {
       await updateProfile(profile, changes);
diff --git a/src/shared/logger.ts b/src/shared/logger.ts
index 27e8259..db15056 100644
--- a/src/shared/logger.ts
+++ b/src/shared/logger.ts
@@ -4,7 +4,11 @@ import { ensureConfigDir, paths } from './config.js';
 
 function formatError(err: unknown): string {
   if (err instanceof Error) return err.stack || err.message || String(err);
-  try { return JSON.stringify(err); } catch { return String(err); }
+  try {
+    return JSON.stringify(err);
+  } catch {
+    return String(err);
+  }
 }
 
 export async function logError(err: unknown, context?: string): Promise<void> {
@@ -25,4 +29,3 @@ export async function logError(err: unknown, context?: string): Promise<void> {
     // swallow all logging errors
   }
 }
-
diff --git a/src/shared/models.ts b/src/shared/models.ts
index b25944d..f3f2177 100644
--- a/src/shared/models.ts
+++ b/src/shared/models.ts
@@ -11,7 +11,11 @@ type Cache = { data: ModelMeta[]; expiresAt: number } | undefined;
 let cache: Cache;
 let inflight: Promise<ModelMeta[]> | null = null;
 
-export async function fetchModelsCached(opts: { domain: string; apiKey?: string; ttlMs?: number }): Promise<ModelMeta[]> {
+export async function fetchModelsCached(opts: {
+  domain: string;
+  apiKey?: string;
+  ttlMs?: number;
+}): Promise<ModelMeta[]> {
   const ttl = Math.max(1000, opts.ttlMs ?? 60_000);
   const now = Date.now();
   if (cache && cache.expiresAt > now) return cache.data;
@@ -30,17 +34,25 @@ export async function fetchModelsCached(opts: { domain: string; apiKey?: string;
 }
 
 export function fuzzyIds(input: string, list: ModelMeta[], limit = 25): string[] {
-  if (!input) return list.slice(0, limit).map(m => m.id);
+  if (!input) return list.slice(0, limit).map((m) => m.id);
   const fuse = new Fuse(list, {
     keys: ['id', 'name'],
     threshold: 0.35,
     ignoreLocation: true,
     includeScore: true,
   });
-  return fuse.search(input).slice(0, limit).map(r => r.item.id);
+  return fuse
+    .search(input)
+    .slice(0, limit)
+    .map((r) => r.item.id);
 }
 
-export function debouncedSuggestFactory(args: { domain: string; apiKey?: string; debounceMs?: number; spinnerLabel?: string }) {
+export function debouncedSuggestFactory(args: {
+  domain: string;
+  apiKey?: string;
+  debounceMs?: number;
+  spinnerLabel?: string;
+}) {
   const debounceMs = Math.max(0, args.debounceMs ?? 200);
   let timer: ReturnType<typeof setTimeout> | null = null;
   let pendingResolve: ((ids: string[]) => void) | null = null;
@@ -72,7 +84,10 @@ export function debouncedSuggestFactory(args: { domain: string; apiKey?: string;
 
     if ((input?.length ?? 0) >= 2 && process.stdout.isTTY) {
       if (!spinner) {
-        spinner = ora({ text: args.spinnerLabel || 'Searching models…', isEnabled: isColorSupported() });
+        spinner = ora({
+          text: args.spinnerLabel || 'Searching models…',
+          isEnabled: isColorSupported(),
+        });
         spinner.start();
       } else if (!spinner.isSpinning) {
         spinner.start();
diff --git a/src/shared/openrouter.ts b/src/shared/openrouter.ts
index 65a1894..b05fd50 100644
--- a/src/shared/openrouter.ts
+++ b/src/shared/openrouter.ts
@@ -1,4 +1,4 @@
-type Message = { role: "system" | "user" | "assistant"; content: string };
+type Message = { role: 'system' | 'user' | 'assistant'; content: string };
 
 export type ChatOptions = {
   domain: string;
@@ -7,11 +7,11 @@ export type ChatOptions = {
   system?: string;
   stream?: boolean;
   onFirstToken?: () => void; // optional UI hook
-  onDone?: () => void;       // optional UI hook
+  onDone?: () => void; // optional UI hook
 };
 
 export async function testConnection({ domain, apiKey }: { domain: string; apiKey: string }) {
-  const url = joinUrl(domain, "models");
+  const url = joinUrl(domain, 'models');
   const res = await fetch(url, {
     headers: {
       Authorization: `Bearer ${apiKey}`,
@@ -24,9 +24,25 @@ export async function testConnection({ domain, apiKey }: { domain: string; apiKe
   return await safeJson(res);
 }
 
-export async function listModels(opts: { domain: string; apiKey?: string }): Promise<{ data: any[] }>
-{
-  const url = joinUrl(opts.domain, "models");
+export async function getCredits({ domain, apiKey }: { domain: string; apiKey: string }) {
+  const url = joinUrl(domain, 'credits');
+  const res = await fetch(url, {
+    headers: {
+      Authorization: `Bearer ${apiKey}`,
+    },
+  });
+  if (!res.ok) {
+    const body = await safeJson(res);
+    throw new Error(`HTTP ${res.status} ${res.statusText}: ${JSON.stringify(body)}`);
+  }
+  return await safeJson(res);
+}
+
+export async function listModels(opts: {
+  domain: string;
+  apiKey?: string;
+}): Promise<{ data: any[] }> {
+  const url = joinUrl(opts.domain, 'models');
   const headers: Record<string, string> = {};
   if (opts.apiKey) headers.Authorization = `Bearer ${opts.apiKey}`;
   const res = await fetch(url, { headers });
@@ -38,17 +54,20 @@ export async function listModels(opts: { domain: string; apiKey?: string }): Pro
   return typeof json === 'object' && json && 'data' in (json as any) ? (json as any) : { data: [] };
 }
 
-export async function askOnce(opts: ChatOptions, messages: Message[]): Promise<string> {
-  const url = joinUrl(opts.domain, "chat/completions");
+export async function askOnce(
+  opts: ChatOptions,
+  messages: Message[]
+): Promise<{ text: string; usage?: any }> {
+  const url = joinUrl(opts.domain, 'chat/completions');
   const body = {
     model: opts.model,
     messages: normalizeMessages(opts, messages),
     stream: false,
   };
   const res = await fetch(url, {
-    method: "POST",
+    method: 'POST',
     headers: {
-      "Content-Type": "application/json",
+      'Content-Type': 'application/json',
       Authorization: `Bearer ${opts.apiKey}`,
     },
     body: JSON.stringify(body),
@@ -59,20 +78,21 @@ export async function askOnce(opts: ChatOptions, messages: Message[]): Promise<s
   }
   const json = await res.json();
   const text: string | undefined = json?.choices?.[0]?.message?.content;
-  return text ?? "";
+  const usage = json?.usage;
+  return { text: text ?? '', usage };
 }
 
 export async function streamChat(opts: ChatOptions, messages: Message[]) {
-  const url = joinUrl(opts.domain, "chat/completions");
+  const url = joinUrl(opts.domain, 'chat/completions');
   const body = {
     model: opts.model,
     messages: normalizeMessages(opts, messages),
     stream: true,
   };
   const res = await fetch(url, {
-    method: "POST",
+    method: 'POST',
     headers: {
-      "Content-Type": "application/json",
+      'Content-Type': 'application/json',
       Authorization: `Bearer ${opts.apiKey}`,
     },
     body: JSON.stringify(body),
@@ -83,26 +103,30 @@ export async function streamChat(opts: ChatOptions, messages: Message[]) {
   }
   const reader = res.body.getReader();
   const decoder = new TextDecoder();
-  let buffer = "";
+  let buffer = '';
   let notified = false;
   while (true) {
     const { done, value } = await reader.read();
     if (done) break;
     buffer += decoder.decode(value, { stream: true });
     let idx;
-    while ((idx = buffer.indexOf("\n\n")) !== -1) {
+    while ((idx = buffer.indexOf('\n\n')) !== -1) {
       const chunk = buffer.slice(0, idx);
       buffer = buffer.slice(idx + 2);
-      for (const line of chunk.split("\n")) {
+      for (const line of chunk.split('\n')) {
         const trimmed = line.trim();
-        if (!trimmed.startsWith("data:")) continue;
+        if (!trimmed.startsWith('data:')) continue;
         const data = trimmed.slice(5).trim();
-        if (data === "[DONE]") return;
+        if (data === '[DONE]') return;
         try {
           const evt = JSON.parse(data);
           const delta: string | undefined = evt?.choices?.[0]?.delta?.content;
           if (delta) {
-            if (!notified && opts.onFirstToken) { try { opts.onFirstToken(); } catch {} }
+            if (!notified && opts.onFirstToken) {
+              try {
+                opts.onFirstToken();
+              } catch {}
+            }
             notified = true;
             process.stdout.write(delta);
           }
@@ -112,21 +136,34 @@ export async function streamChat(opts: ChatOptions, messages: Message[]) {
   }
   if (buffer.length) {
     try {
-      const evt = JSON.parse(buffer.replace(/^data:\s*/, ""));
+      const evt = JSON.parse(buffer.replace(/^data:\s*/, ''));
       const delta: string | undefined = evt?.choices?.[0]?.delta?.content;
       if (delta) {
-        if (!notified && opts.onFirstToken) { try { opts.onFirstToken(); } catch {} }
+        if (!notified && opts.onFirstToken) {
+          try {
+            opts.onFirstToken();
+          } catch {}
+        }
         notified = true;
         process.stdout.write(delta);
       }
     } catch {}
   }
-  if (opts.onDone) { try { opts.onDone(); } catch {} }
+  if (opts.onDone) {
+    try {
+      opts.onDone();
+    } catch {}
+  }
+  if (opts.onDone) {
+    try {
+      opts.onDone();
+    } catch {}
+  }
 }
 
 function normalizeMessages(opts: ChatOptions, messages: Message[]) {
   const out: Message[] = [];
-  if (opts.system) out.push({ role: "system", content: opts.system });
+  if (opts.system) out.push({ role: 'system', content: opts.system });
   out.push(...messages);
   return out;
 }
@@ -141,8 +178,8 @@ async function safeJson(res: Response) {
 
 export function joinUrl(base: string, path: string) {
   const u = new URL(base);
-  const basePath = u.pathname.replace(/\/+$/, "");
-  const rel = path.replace(/^\/+/, "");
+  const basePath = u.pathname.replace(/\/+$/, '');
+  const rel = path.replace(/^\/+/, '');
   u.pathname = `${basePath}/${rel}`;
   return u;
 }
diff --git a/src/shared/ui.ts b/src/shared/ui.ts
index 9789730..0695d8c 100644
--- a/src/shared/ui.ts
+++ b/src/shared/ui.ts
@@ -45,7 +45,7 @@ export function examplesBox(): string {
     ['test', 'openrouter test'],
   ];
   const header = `${palette.bold('Examples')}`;
-  const col1Width = Math.max(...rows.map(r => r[0].length), 'Command'.length) + 2;
+  const col1Width = Math.max(...rows.map((r) => r[0].length), 'Command'.length) + 2;
   const lines = [
     `${pad('Command', col1Width)}Example`,
     `${'-'.repeat(col1Width)}${'-'.repeat(32)}`,
@@ -95,7 +95,8 @@ export function styledPrompt(model: string): string {
 
 export function tipBox(): string {
   const lines = [
-    palette.dim("Type 'exit' to quit. Commands: ") + '/model <name>, /system <text>, /format <md|plain>, /stream <on|off>',
+    palette.dim("Type 'exit' to quit. Commands: ") +
+      '/model <name>, /system <text>, /format <md|plain>, /stream <on|off>, /stats, /billing',
     palette.dim("Use 'openrouter init' to change defaults."),
   ].join('\n');
   return boxen(lines, { padding: 1, borderStyle: 'single' });
@@ -103,6 +104,16 @@ export function tipBox(): string {
 
 export function showSpinner(label: string) {
   const enabled = isColorSupported();
-  const spinner = ora({ text: palette.dim(label), isEnabled: enabled, stream: process.stderr as any });
+  const spinner = ora({
+    text: palette.dim(label),
+    isEnabled: enabled,
+    stream: process.stderr as any,
+  });
   return spinner;
 }
+
+export function warnBox(text: string): string {
+  // Emphasize with red text and a clear border; remains readable without color
+  const body = palette.err(text);
+  return boxen(body, { padding: 1, borderStyle: 'round' });
+}
diff --git a/tests/config-redaction.spec.ts b/tests/config-redaction.spec.ts
index f28abe7..f141a8a 100644
--- a/tests/config-redaction.spec.ts
+++ b/tests/config-redaction.spec.ts
@@ -38,7 +38,9 @@ describe('config --list redaction', () => {
     const program = main.buildProgram();
     let out = '';
     const origLog = console.log;
-    console.log = (s?: any) => { out += (typeof s === 'string' ? s : String(s)); };
+    console.log = (s?: any) => {
+      out += typeof s === 'string' ? s : String(s);
+    };
     try {
       await program.parseAsync(['node', 'cli', 'config', '--list']);
     } finally {
diff --git a/tests/config-yaml.spec.ts b/tests/config-yaml.spec.ts
index 6e4dc43..64daa0c 100644
--- a/tests/config-yaml.spec.ts
+++ b/tests/config-yaml.spec.ts
@@ -31,7 +31,11 @@ afterEach(async () => {
 describe('project YAML rc precedence', () => {
   it('reads .openrouterrc.yaml and overrides model', async () => {
     await fs.mkdir(cfgModule.paths.CONFIG_DIR, { recursive: true });
-    await cfgModule.updateConfig({ domain: 'https://global.example/v1', model: 'global-model', apiKey: 'sk-global' });
+    await cfgModule.updateConfig({
+      domain: 'https://global.example/v1',
+      model: 'global-model',
+      apiKey: 'sk-global',
+    });
     await cfgModule.updateProfile('dev', { domain: 'https://dev.example/v1', model: 'dev-model' });
 
     const rcPath = path.join(process.cwd(), '.openrouterrc.yaml');
@@ -43,4 +47,3 @@ describe('project YAML rc precedence', () => {
     expect(eff.apiKey).toBe('sk-global');
   });
 });
-
diff --git a/tests/config.spec.ts b/tests/config.spec.ts
index b27ea23..d376cce 100644
--- a/tests/config.spec.ts
+++ b/tests/config.spec.ts
@@ -39,7 +39,11 @@ describe('config precedence and profiles', () => {
   it('merges global base + profile + project rc (project wins)', async () => {
     // Write global base config
     await fs.mkdir(cfgModule.paths.CONFIG_DIR, { recursive: true });
-    await cfgModule.updateConfig({ domain: 'https://global.example/v1', model: 'global-model', apiKey: 'sk-global' });
+    await cfgModule.updateConfig({
+      domain: 'https://global.example/v1',
+      model: 'global-model',
+      apiKey: 'sk-global',
+    });
     await cfgModule.updateProfile('dev', { domain: 'https://dev.example/v1', model: 'dev-model' });
 
     // Write project rc overriding model only
diff --git a/tests/format.spec.ts b/tests/format.spec.ts
index 0e419a0..07f34de 100644
--- a/tests/format.spec.ts
+++ b/tests/format.spec.ts
@@ -13,4 +13,3 @@ describe('markdown rendering (ANSI)', () => {
     expect(out).toMatch(/\x1b\[2mcode\x1b\[22m/);
   });
 });
-
diff --git a/tests/help.spec.ts b/tests/help.spec.ts
index 5c9799d..bdce7b5 100644
--- a/tests/help.spec.ts
+++ b/tests/help.spec.ts
@@ -4,24 +4,23 @@ import { buildProgram } from '../src/main.js';
 describe('CLI help output', () => {
   it('lists core commands and omits ask model override', () => {
     const program = buildProgram();
-    const names = program.commands.map(c => c.name());
+    const names = program.commands.map((c) => c.name());
     expect(names).toEqual(expect.arrayContaining(['init', 'config', 'ask', 'repl', 'test']));
 
-    const ask = program.commands.find(c => c.name() === 'ask')!;
+    const ask = program.commands.find((c) => c.name() === 'ask')!;
     const askHelp = ask.helpInformation();
     expect(askHelp).not.toContain('--model');
     expect(askHelp).toContain('--format');
     expect(askHelp).toContain("Use 'openrouter init' to change defaults");
 
-    const cfg = program.commands.find(c => c.name() === 'config')!;
+    const cfg = program.commands.find((c) => c.name() === 'config')!;
     const cfgHelp = cfg.helpInformation();
     expect(cfgHelp).toContain('--api-key');
     expect(cfgHelp).not.toMatch(/--domain|--model|--provider/);
 
-    const repl = program.commands.find(c => c.name() === 'repl')!;
+    const repl = program.commands.find((c) => c.name() === 'repl')!;
     const replHelp = repl.helpInformation();
     expect(replHelp).toContain('/format');
     expect(replHelp).toContain('/stream');
   });
 });
-
diff --git a/tests/models.spec.ts b/tests/models.spec.ts
index 6479b15..b3f36a6 100644
--- a/tests/models.spec.ts
+++ b/tests/models.spec.ts
@@ -4,12 +4,14 @@ vi.mock('../src/shared/openrouter.js', async (orig) => {
   const actual = await (orig as any)();
   return {
     ...actual,
-    listModels: vi.fn(async () => ({ data: [
-      { id: 'meta-llama/llama-3.1-8b-instruct', name: 'Llama 3.1 8B Instruct' },
-      { id: 'openai/gpt-4o-mini', name: 'GPT-4o Mini' },
-      { id: 'google/gemini-pro', name: 'Gemini Pro' },
-      { id: 'microsoft/phi-3-mini', name: 'Phi-3 Mini' },
-    ] }))
+    listModels: vi.fn(async () => ({
+      data: [
+        { id: 'meta-llama/llama-3.1-8b-instruct', name: 'Llama 3.1 8B Instruct' },
+        { id: 'openai/gpt-4o-mini', name: 'GPT-4o Mini' },
+        { id: 'google/gemini-pro', name: 'Gemini Pro' },
+        { id: 'microsoft/phi-3-mini', name: 'Phi-3 Mini' },
+      ],
+    })),
   };
 });
 
@@ -34,19 +36,26 @@ describe('models helper', () => {
 
   it('fetchModelsCached caches results by ttl', async () => {
     const first = await fetchModelsCached({ domain: 'https://openrouter.ai/api/v1', ttlMs: 60000 });
-    const second = await fetchModelsCached({ domain: 'https://openrouter.ai/api/v1', ttlMs: 60000 });
+    const second = await fetchModelsCached({
+      domain: 'https://openrouter.ai/api/v1',
+      ttlMs: 60000,
+    });
     expect(second).toBe(first); // same array instance implies cache hit
   });
 
   it('debouncedSuggestFactory falls back to default when fetch fails', async () => {
     // Override mock to throw
     const { listModels } = await import('../src/shared/openrouter.js');
-    (listModels as any).mockImplementationOnce(async () => { throw new Error('offline'); });
-    const suggest = debouncedSuggestFactory({ domain: 'https://openrouter.ai/api/v1', debounceMs: 200 });
+    (listModels as any).mockImplementationOnce(async () => {
+      throw new Error('offline');
+    });
+    const suggest = debouncedSuggestFactory({
+      domain: 'https://openrouter.ai/api/v1',
+      debounceMs: 200,
+    });
     const p = suggest('ll');
     vi.advanceTimersByTime(210);
     const out = await p;
     expect(out.length).toBeGreaterThan(0);
   });
 });
-
diff --git a/tests/openrouter-ask-error.spec.ts b/tests/openrouter-ask-error.spec.ts
index fd39cf3..72740b7 100644
--- a/tests/openrouter-ask-error.spec.ts
+++ b/tests/openrouter-ask-error.spec.ts
@@ -16,7 +16,7 @@ describe('askOnce error handling', () => {
       ok: false,
       status: 400,
       statusText: 'Bad Request',
-      json: async () => ({ error: 'test-error' })
+      json: async () => ({ error: 'test-error' }),
     })) as any;
 
     const opts: ChatOptions = {
@@ -25,8 +25,8 @@ describe('askOnce error handling', () => {
       model: 'm',
       stream: false,
     };
-    await expect(askOnce(opts, [{ role: 'user', content: 'hi' }]))
-      .rejects.toThrow(/HTTP 400 Bad Request: .*test-error.*/);
+    await expect(askOnce(opts, [{ role: 'user', content: 'hi' }])).rejects.toThrow(
+      /HTTP 400 Bad Request: .*test-error.*/
+    );
   });
 });
-
diff --git a/tests/openrouter-stream.spec.ts b/tests/openrouter-stream.spec.ts
index 496df73..7efebcd 100644
--- a/tests/openrouter-stream.spec.ts
+++ b/tests/openrouter-stream.spec.ts
@@ -15,8 +15,8 @@ describe('streamChat SSE parsing', () => {
   it('writes deltas to stdout for chunks and stops at [DONE]', async () => {
     const chunks = [
       'data: {"choices":[{"delta":{"content":"Hello "}}]}\n\n' +
-      'data: {"choices":[{"delta":{"content":"world"}}]}\n\n',
-      'data: [DONE]\n\n'
+        'data: {"choices":[{"delta":{"content":"world"}}]}\n\n',
+      'data: [DONE]\n\n',
     ];
     const reader = {
       i: 0,
@@ -26,11 +26,11 @@ describe('streamChat SSE parsing', () => {
           return { done: false, value: v };
         }
         return { done: true };
-      }
+      },
     };
     globalThis.fetch = (async () => ({
       ok: true,
-      body: { getReader: () => reader }
+      body: { getReader: () => reader },
     })) as any;
 
     const writes: string[] = [];
@@ -42,7 +42,12 @@ describe('streamChat SSE parsing', () => {
     };
 
     try {
-      const opts: ChatOptions = { domain: 'https://example.com/v1', apiKey: 'sk', model: 'm', stream: true };
+      const opts: ChatOptions = {
+        domain: 'https://example.com/v1',
+        apiKey: 'sk',
+        model: 'm',
+        stream: true,
+      };
       await streamChat(opts, [{ role: 'user', content: 'hi' }]);
     } finally {
       process.stdout.write = origWrite;
diff --git a/tests/openrouter.spec.ts b/tests/openrouter.spec.ts
index 90f3ce9..b405cc4 100644
--- a/tests/openrouter.spec.ts
+++ b/tests/openrouter.spec.ts
@@ -3,13 +3,14 @@ import { joinUrl } from '../src/shared/openrouter.js';
 
 describe('joinUrl', () => {
   it('preserves base path when joining', () => {
-    expect(joinUrl('https://openrouter.ai/api/v1', 'models').toString())
-      .toBe('https://openrouter.ai/api/v1/models');
+    expect(joinUrl('https://openrouter.ai/api/v1', 'models').toString()).toBe(
+      'https://openrouter.ai/api/v1/models'
+    );
   });
 
   it('handles trailing and leading slashes correctly', () => {
-    expect(joinUrl('https://example.com/root/', '/chat/completions').toString())
-      .toBe('https://example.com/root/chat/completions');
+    expect(joinUrl('https://example.com/root/', '/chat/completions').toString()).toBe(
+      'https://example.com/root/chat/completions'
+    );
   });
 });
-
diff --git a/tests/ui-answer.spec.ts b/tests/ui-answer.spec.ts
index 8936413..2922a11 100644
--- a/tests/ui-answer.spec.ts
+++ b/tests/ui-answer.spec.ts
@@ -14,4 +14,3 @@ describe('answer render blocks', () => {
     expect(s).toContain('openrouter.ai');
   });
 });
-
diff --git a/vitest.config.ts b/vitest.config.ts
index 891499c..19a0739 100644
--- a/vitest.config.ts
+++ b/vitest.config.ts
@@ -10,4 +10,3 @@ export default defineConfig({
     },
   },
 });
-