Unofficial command-line interface for the Brainville API v2. Not affiliated with or endorsed by Brainville AB. API docs: https://developer.brainville.com/api/v2/.
A single static binary (no runtime needed) for people and AI agents: search marketplace assignments and manage your organisation's own assignments, consultants and applications. Includes an agent skill (skills/konsult/).
Download the archive for your platform from GitHub Releases, check it against checksums.txt, and put konsult on your PATH. From v0.1.1 the archives carry a build provenance attestation: gh attestation verify konsult_darwin_arm64.tar.gz --repo activout/konsult-cli. Or build: go build -o konsult ./cmd/konsult.
Prerequisite: Brainville API access (user key + sender key). Market needs Premium Sales, applications need Premium Buyer or the Applications API licence.
konsult auth login # prompts without echo; stores the keys in the OS keychain
konsult auth status --verify # never prints the keys
konsult auth logout- Storage: macOS Keychain, Windows Credential Manager, Linux Secret Service. Keys are never accepted as arguments.
- Piped:
printf '%s\n%s\n' "$USER_KEY" "$SENDER_KEY" | konsult auth login(user key first). - Headless/CI: set
KONSULT_USER_KEYandKONSULT_SENDER_KEY; they take precedence over the keychain. - Precedence: flag > environment variable > keychain > default. Default environment is production. For Brainville's test environment use
--env sandbox(orKONSULT_ENV=sandbox); keys are stored per environment.
konsult consultants list --ndjson
konsult assignments list --json
konsult assignments get 12345
konsult assignments create --file draft.json --dry-run
konsult assignments delete 12345 --yes
konsult market search --text 'C# .NET' --match and --ndjson # consumes export quota
konsult applications list --assignment 12345 --ndjson
konsult applications resume 777 -o cv.pdf
konsult completion zshOutput: table on a terminal, JSON when piped, --json (service JSON) or --ndjson (one compact record per line; market search emits one line per result). In those modes stdout holds only JSON; messages go to stderr.
Safety: every change has --dry-run (redacted payload), asks for confirmation, and refuses to run non-interactively without --yes. boost is paid and needs --accept-charge. close rejects pending applications and emails applicants. Downloads never overwrite without --force and are created with mode 0600.
Exit codes: 0 ok, 1 general, 2 usage, 3 auth/licence, 4 not found, 5 invalid input, 6 rate limited, 7 server error, 8 not confirmed.
Copy or symlink skills/konsult into your agent's skills directory (for Claude Code: ~/.claude/skills/konsult). It contains no secrets.
See docs/api-coverage.md, docs/sandbox-findings.md, docs/open-questions.md and openapi/unofficial-v2.yaml (community-authored, evidence-tagged). Sandbox-verified: consultants, assignments, files, market search/get, unpublish/close/delete/share. Not verified: applications (licence), boost on (404 in the sandbox), and whether emails are sent. Not covered: webhooks, v1, XML. The macOS keychain backend runs /usr/bin/security.
Troubleshooting: HTTP 401 on applications/market usually means a missing licence rather than a wrong key. --debug logs method, path and status only.
Use requires your company's approved Brainville API access and licences. Brainville's Terms of Use allow automated tools, including AI agents, only through authorised APIs, and marketplace data may only be used inside your own legal entity. See docs/terms-review.md (a summary, not legal advice).
See SECURITY.md. Licence: MIT (see LICENSE).