delorenj/mcp-server-trello

MCP Server Trello – AI‑friendly Trello bridge

What it is – A tiny, Bun‑powered server that implements the Model Context Protocol (MCP). It wraps the Trello REST API and exposes 57 typed tools (e.g., get_card, add_checklist_item, watch_card) that AI agents can call as if they were native functions. The server handles authentication, rate‑limiting, input validation and returns data in JSON or ready‑to‑paste markdown.

Why it matters – Many AI‑agent frameworks (Claude Desktop, Cursor, etc.) need a reliable way to read/write a team’s project board. This server gives agents full board access without writing custom HTTP code and adds higher‑level helpers such as:

• Acceptance‑criteria extraction (get_acceptance_criteria) that automatically finds the right checklist (AC, DoD, etc.) and reports completion percentage.
• Watch tools (watch_card, watch_list) that push Trello activity into the agent’s notification stream.
• Dynamic board/workspace switching so the same server can serve many projects without a restart.

Key features

• Complete CRUD for cards, lists, boards, checklists, comments, attachments, labels, custom fields.
• Built‑in rate‑limit handling (300 req/10 s per API key, 100 req/10 s per token).
• Type‑safe TypeScript implementation – agents get proper argument schemas.
• Optional environment‑variable sandboxing (TRELLO_ALLOWED_WORKSPACES) for multi‑tenant security.
• Markdown export of card data for human‑readable context.
• Distributed as a single npm package (@delorenj/mcp-server-trello) that can be launched with bunx or npx.

How to get it running

1. Install – npm i -g @delorenj/mcp-server-trello (or just reference the package in your MCP client; no cloning needed).
2. Configure – create a .env with TRELLO_API_KEY and TRELLO_TOKEN (obtainable from https://trello.com/app-key). Optional vars let you set a default board, workspace, proxy, or restrict allowed workspaces.
3. Add to an MCP client – e.g. in Claude Desktop’s mcpServers config:

   {
     "trello": {
       "command": "bunx",
       "args": ["@delorenj/mcp-server-trello"],
       "env": {"TRELLO_API_KEY": "…", "TRELLO_TOKEN": "…"}
     }
   }
   

4. Start – bunx @delorenj/mcp-server-trello (or npx …). The server registers itself with the MCP registry, so any registry‑aware client can discover it automatically.
5. Use the tools – Agents call the exposed tools by name with the JSON payload shown in the README (e.g., { "name": "get_card", "arguments": { "cardId": "abc123" } }). Responses include rich data structures and, when requested, markdown rendering.

Skill package – The repo also ships a BMAD‑compatible skill directory. Installing the skill makes the server’s tools appear in an agent’s skill set, and the provided install.sh script builds the server locally if Bun is present.

Typical workflow

1. list_boards → pick a board ID
2. set_active_board (or set_active_workspace)
3. get_card / get_checklist_items / get_acceptance_criteria
4. update_checklist_item / move_card / add_comment …
5. watch_card to get live updates

Who should use it – Teams building AI assistants that need to read or modify Trello boards (project‑management bots, sprint‑planning agents, automated retrospectives) and anyone who wants a typed, rate‑limited, MCP‑compliant interface instead of raw HTTP calls.

---

Links

• npm: https://www.npmjs.com/package/@delorenj/mcp-server-trello
• MCP Registry entry: https://registry.modelcontextprotocol.io/servers/io.github.delorenj/mcp-server-trello
• Demo badge: https://glama.ai/mcp/servers/klqkamy7wt