MCP integration
Connect 1VEC to Claude Code and other agents so AI can read and write issues directly.
1VEC ships a built-in MCP server that lets Claude Code and other agents treat issues and projects as first-class objects to read and write directly.
Built on the public data plane
The MCP server reaches 1VEC through the public data plane (/api/public/v1/*) with a scoped API key — never the internal API. That way third-party agents and your own use the same stable public surface.
- It holds a scoped key (bound to an org + authorized projects + a role), not super; the key decides which orgs / projects it can touch.
- Multiple environments = multiple named MCP instances, each holding its environment's key (production keys are managed and revoked separately).
Add it to Claude Code
cd apps/mcp
cp .env.template .env # set ONEVEC_API_URL (public plane) + ONEVEC_API_KEY (scoped)
claude mcp add 1vec -- bun run <repo>/apps/mcp/src/index.tsTransports
- stdio — local Claude Code mounting (common in dev; usually not deployed).
- HTTP / SSE — enable it for multi-machine or remote sharing.
Available tools
Only capabilities that exist on the public plane are exposed:
orgs / projectslistissueslist / get / create / update / commentworkflows…
The tool list evolves with the public plane; see
tools/liston each environment's MCP instance for details.