Model Context Protocol (MCP) server for Kanboard — manage projects, tasks, columns, swimlanes, and team work from any MCP-compatible client (Claude Code, Claude Desktop, Cursor, Cline, Zed, and more).
claude mcp add kanboard-mcp -- npx -y kanboard-mcp
A pasteable skeleton generated from the repository name. Check the project README for the exact package and arguments.
Star count
9 ★ · ErnestoCorona/kanboard-mcp · read from the GitHub search API · (checked 2026-10-16 20:00:20)
Source
collected from the punkpeye/awesome-mcp-servers list · type: MCP server
Harvest history
first seen 2026-10-09 · last seen 2026-10-16 · seen 32× · listed by 7 sources
Score
score 703 · 9 ★ · 7 sources · mcp
FAQ (answers come from our records)
How do I install it?
We did not find an install command for this repository (our install coverage is only about 4%). Check its README or releases page.
Who is it for?
We filed it under these tags: MCP 服务. The tags come from the entry itself, not a taxonomy we imposed.
Who maintains it?
repository ErnestoCorona/kanboard-mcp;first catalogued 2026-10-09;last seen 2026-10-16。There is no official certification or quality rating here — we record sources and dates only.
Kanboard, plug-and-play in any AI editor. A Model Context Protocol server that brings your Kanboard board into Claude Code, Claude Desktop, Cursor, Cline, Zed, and beyond — so your agent can read, plan, and update tasks the same way you would. **39 typed tools · JSON-RPC batching · Dual authentication · TypeScript strict · 1016 tests** --- What it looks like The same flow works for every kind of board work — pulling overdue tasks, opening a daily standup summary, breaking a doc into subtasks, or updating a comment on someone's PR. The agent picks the tools, you stay in the loop. --- Why Kanboard MCP - **39 typed tools** across 9 groups — full CRUD on projects, columns, swimlanes, tasks, subtasks, comments, attachments, and members. No half-supported entities, no read-only stubs. - **JSON-RPC 2.0 batching** — create up to 100 tasks in one HTTP round-trip. Turn an email, a meeting transcript, or a doc into a sprint backlog in seconds. - **Dual authentication** — *personal token* (acts as you, with your Kanboard identity) or *application token* (service identity for CI, bots, and shared agents). - **Walk-up project resolver** — drop one .kanboard.yaml at your repo root, every tool auto-resolves the project context. Switch repos, your agent switches boards. - **Zero runtime HTTP dependencies** — native Node 22 fetch, 4 production deps total. Audit surface is intentionally tiny. - **Pino structured logging with automatic secret redaction** — token values never reach stdout, stderr, or any log line, at any log level. - **TypeScript strict mode** + **1016 unit tests** across 65 test files. Integration suite gated against accidental writes to non-sandbox projects. - **Smart retries for reads only** — idempotent calls retry transparently on transient HTTP failures (429 / 502 / 503 / 504); mutations never retry. - **Hard per-request timeouts** — every JSON-RPC call runs under AbortSignal.timeout() (default 15 s, configurable via KANBOARD_TIMEOUT_MS). Requests cannot hang the agent indefinitely — slow or unresponsive backends surface as a clean TimeoutError your agent can recover from. - **Debuggable from day one** — speaks plain MCP over stdio, so the official MCP Inspector works out of the box. Inspect schemas, fire individual tool calls, watch JSON-RPC traffic in a browser UI. See Debugging with MCP Inspector. 60-second quick start 1. Get a Kanboard API token In Kanboard: **Profile → API → Generate token** (personal mode), or **Settings → API → Application token** (app mode for service accounts). 2. Add Kanboard MCP to your client **Recommended path: npx — zero install, always latest.** Drop this block into your MCP client config: Restart your MCP client. Done — the 39 tools are now available to your agent. > **Why npx?** It fetches @ernestocorona/kanboard-mcp from npm on demand, caches it locally, and runs it as the MCP server. You never run a separate install command, and you always get the latest published version — no upgrade chore, no $PATH to manage. This is also why step 3's selftest works without any prior install: npx handles the fetch transparently. 3. Verify Expected output (exit 0 = ready): Documentation Full docs live in ./docs/ and follow the Diátaxis framework: - **Tutorials** — hand-held walkthroughs to learn by doing. - **How-to guides** — recipes for batching, multi-project setups, integration tests, and CI. - **Reference** — exact contracts for every tool, every config knob, and every error. - **Explanation** — the *why* behind authentication modes, the retry policy, and the batch architecture. Start at the docs index to pick the right entry point. Installation methods **Just want it working?** Use the npx row (first entry below) — it's the recommended path for the vast majority of users. The other methods exist for specific use cases (frequent local runs, Bun/pnpm runtimes, production deployments, or hacking on the source). | Method | When to use | Command | |--------|-------------|---------| | **npx** *(recommended)* | Most users. Zero install, always uses the latest published version. | npx -y @ernestocorona/kanboard-mcp | | **Global install** | You run the server frequently and want a stable binary on $PATH. | npm i -g @ernestocorona/kanboard-mcp then kanboard-mcp |