Open Agent Skills and plugins for AI coding agents working with the Spice.ai OSS runtime — data federation, acceleration, search, AI/LLM, and cloud management.
This is the Spice.ai Marketplace: packaged skills/plugins any compatible harness can load. The format works across Claude Code, Cursor, Codex, Grok, OpenCode, Pi, and other agents that support the open Agent Skills / plugin standard — not Claude-only.
Ask your agent to set up Spice, for example "set up Spice in ./my_app and show me a SQL result".
The setup skill is the entry point. It checks your existing installation, creates a
spicepod with a local sample dataset (or your own file), starts the runtime, waits for /v1/ready,
and finishes only when a query returns rows. Then it hands off to the specialists:
setup → spicepod, secrets → connectors + datasets → acceleration, accelerators, cache, search, models, chat, sql, sdk
To go further — a scenario such as "serve our Snowflake, Postgres, and S3 data to agents with OpenAI"
deployed to Spice.ai Cloud, tested end to end, monitored with alerts, and handed off with a runbook —
use launch. cookbook runs ready-made recipes instead of your own
project, and cloud and terraform manage Spice.ai Cloud resources. You install Spice
itself by following the installation docs; the skills check
spice version and never install or upgrade it.
| Agent | Install |
|---|---|
| Claude Code | /plugin marketplace add spiceai/skills, then /plugin install spiceai@spiceai |
| OpenAI Codex | codex plugin marketplace add spiceai/skills, then codex plugin add spiceai@spiceai |
| GitHub Copilot | copilot plugin marketplace add spiceai/skills, then copilot plugin install spiceai@spiceai |
| Grok Build | grok plugin marketplace add spiceai/skills, then grok plugin install spiceai --trust |
| Cursor, Grok Bot, OpenCode, Pi, others | npx skills add spiceai/skills -a <agent> (cursor, opencode, pi, ...) |
Plugin installs invoke skills as /spiceai:setup (Claude Code) or $spiceai:setup (Codex);
npx skills add installs them without the namespace ($setup). Plugin installs also add the
Spice.ai Cloud management MCP server as spice-cloud (https://api.spice.ai/mcp); sign in from
the agent's MCP menu (/mcp in Claude Code). The sections below cover global installs, project
auto-discovery, and maintainer publishing.
Per-agent installation details and maintainer publishing notes
The plugin identifier is spiceai. When installed as a plugin, invoke its skills
as $spiceai:setup, $spiceai:chat, or $spiceai:spicepod.
For local plugin testing from this checkout:
codex plugin marketplace add .
codex plugin add spiceai@spiceaiThe commands below install individual skills without a plugin namespace:
npx skills add spiceai/skills -a codexGlobal (all projects):
npx skills add spiceai/skills -g -a codexSkills install under .agents/skills/ (project) or ~/.codex/skills/ (global). Inside a Codex session you can also run $skill-installer and point it at spiceai/skills, then restart Codex if the skills do not appear. Verify with /skills.
See Codex Skills.
Publish (maintainers): build the skills-only upload package:
make check-versions
make packageIn the OpenAI Plugins dashboard, select the owning organization and project, choose Upload new or existing plugin, select the verified Spice AI developer identity, and upload dist/spiceai-plugin-<version>.zip. Resolve the metadata and skill scan findings, submit for review, and select Publish plugin after approval. Upload subsequent ZIPs to the existing listing using the same package name. If a draft was created under an older package name, create a new spiceai draft: OpenAI rejects package-name changes on an existing listing. See OpenAI's submission guide for publishing permissions and verification requirements.
Listing metadata and bundled icons come from plugin.json and assets/, with matching metadata in the Codex compatibility manifest. The local/repo catalog (codex plugin marketplace add spiceai/skills) supports local testing and distribution; adding it does not publish to the public directory. GitHub releases attach both archives, checksums, and marketplace submission materials using the same packaging script. See the publishing runbook.
copilot plugin marketplace add spiceai/skills
copilot plugin install spiceai@spiceaiCopilot uses .github/plugin/marketplace.json and the portable
root plugin.json. For a curated listing, contribute the release's generated catalog
entry to github/copilot-plugins, following its contribution guide.
npx skills add spiceai/skills -a grokOr install as a Grok plugin from the repo:
grok plugin install spiceai/skills --trustAdd this repo as a marketplace source, then install spiceai:
grok plugin marketplace add spiceai/skills
grok plugin install spiceai --trustBrowse/install from the TUI with /marketplace or /plugins. Skills land in .grok/skills/ (project) or ~/.grok/skills/ (global). Native manifests: .grok-plugin/plugin.json + .grok-plugin/marketplace.json. Grok also reads Claude Code marketplaces and .agents/skills/ with no extra setup — see Skills, Plugins & Marketplaces.
Publish (maintainers): after merge, open a PR to xai-org/plugin-marketplace adding a remote catalog entry for spiceai pinned to a full commit sha of spiceai/skills (see their CONTRIBUTING). Until that lands, users can still grok plugin marketplace add spiceai/skills or grok plugin install spiceai/skills --trust.
Grok Bot teammates use the same Agent Skills format. Install Spice skills into the Cursor skill paths the Bot shares:
npx skills add spiceai/skills -a cursorOr ask a Grok Bot (or open Settings → Plugins) to add the spiceai/skills marketplace/plugin. After install, start a new Bot turn so the skills catalog refreshes.
Install skills into Cursor paths:
npx skills add spiceai/skills -a cursorOr add this repo as a Cursor plugin / Team Marketplace source (requires .cursor-plugin/plugin.json).
Publish (maintainers): submit https://github.com/spiceai/skills at cursor.com/marketplace/publish after manifests land on trunk. Plugin id: spiceai @ 2.3.2.
npx skills add spiceai/skillsTarget a specific harness with -a (examples: opencode, pi, claude-code, cursor). Use -g for a user-global install. Once installed, skills activate when the agent detects a relevant Spice task.
/plugin marketplace add spiceai/skills
/plugin install spiceai@spiceai
Skills are then available as /spiceai:setup, /spiceai:chat, /spiceai:spicepod, etc.
For the public Claude directory, submit this GitHub repository through the directory submission portal, then complete review and publication. Registering the repository marketplace is independent of directory approval.
To auto-suggest the plugin for all contributors, add this to your project's .claude/settings.json:
{
"extraKnownMarketplaces": {
"spiceai": {
"source": {
"source": "github",
"repo": "spiceai/skills"
}
}
},
"enabledPlugins": {
"spiceai@spiceai": true
}
}Pull requests validate metadata, private-skill exclusion, and packaging, and produce a submission
preview. Pushing a v<version> tag creates a draft GitHub release containing ZIP/tar archives,
checksums, full-SHA catalog entries for Claude/Copilot/Grok, and an OpenAI upload guide. Publishing
a GitHub release or manually rerunning the release workflow refreshes those assets and the generated
release-note section while preserving authored notes. Claude/OpenAI portal review and GitHub/Grok
catalog PRs remain separate publishing steps. See the publishing runbook.
Skills versions match the Spice.ai OSS runtime they target (e.g. 2.3.2 with runtime v2.3.2). Pin to GitHub release tag v2.3.2 when you need a fixed surface, or use trunk for latest.
The skills are version-aware. Each one names the release line it is written for, tells the agent to check your runtime version (spice version) before it recommends configuration, marks features added after v2.0.0 with the release that shipped them (for example (v2.2.0+)), and lists removed or deprecated configuration with its replacement. On an older runtime, the agent can then avoid newer settings and use the docs for your release (https://spiceai.org/docs/v2.2/...).
| Skill | Description |
|---|---|
| setup | Start here. Take a directory to a running runtime that answers SQL with rows, using an existing installation; CLI, ports, and /v1/sql reference |
| launch | Sign in (or up) to Spice.ai Cloud with a device code, then take a scenario to a Cloud project that is deployed, verified (rows, model answers, MCP), monitored with alerts and a fire drill, and handed off with a runbook; for demos, POCs, and greenfield production |
| cookbook | Find, set up, and run recipes from the Spice.ai cookbook, including from a pull request |
| spicepod | Create and configure spicepod.yaml manifests |
| datasets | Connect to data sources and query across them with federated SQL |
| connectors | Configure individual data source connectors (PostgreSQL, S3, Snowflake, etc.) |
| acceleration | Accelerate data locally for sub-second query performance |
| accelerators | Choose and configure acceleration engines (Arrow, DuckDB, SQLite, etc.) |
| search | Search with vector similarity, full-text keywords, or hybrid RRF |
| chat | Chat through the CLI or API, with tools, memory, and model routing |
| models | Configure LLM providers (OpenAI, Anthropic, Azure, local GGUF, etc.) |
| sql | Author and run SQL through the CLI, runtime API, or SDK; build text-to-SQL workflows |
| cache | Cache query and search results with TTL and stale-while-revalidate |
| secrets | Manage credentials with secret stores |
| cloud | Manage Spice.ai Cloud resources via the Management API and its MCP server |
| terraform | Manage Spice.ai Cloud infrastructure as code with Terraform |
| sdk | Integrate Python, JavaScript/TypeScript, Go, Rust, Java, and .NET applications |
The skill names follow the CLI and API concepts where they overlap. A skill can cover several related commands; configuration topics are not invented CLI subcommands.
| Skill | CLI, API, or SDK surface |
|---|---|
setup |
Existing installation, spice init, spice validate, spice run, spice version, /v1/ready |
sql |
spice sql, /v1/sql, /v1/nsql, SDK SQL methods |
search |
spice search, /v1/search, SDK search methods |
chat / models |
spice chat, spice models, OpenAI-compatible APIs and provider configuration |
datasets / connectors |
Dataset and catalog configuration, spice datasets, /v1/datasets |
cloud |
Available spice cloud commands, the Cloud Management API, and the management MCP server (https://api.spice.ai/mcp) |
launch |
spice cloud login/project/deploy/status/logs, Management API monitors, forks, org-secret links, and pause, /v1/sql, /v1/chat/completions, /v1/mcp |
sdk |
Client construction, endpoints, credentials, typed results, and language-specific methods |
spicepod, acceleration, accelerators, cache, secrets |
Runtime configuration and related operations |
terraform / cookbook |
Infrastructure as code and documented recipe workflows |
Cloud runtime calls use project endpoints and project API keys. Cloud management operations
use the Management API and its access tokens. SDK versions are checked independently of the
runtime version. Installing individual skills without the plugin uses their short names,
such as $sql; plugin installs use $spiceai:sql or /spiceai:sql in the corresponding agent.
- Spice.ai OSS GitHub
- Spice Documentation
- Spicepod Reference
- Cookbook
- Introducing Spice Skills
- Agent Skills CLI
MIT