Read-only Trello MCP server
Lets Claude read your Trello boards, lists, and cards via a remote MCP server.
README
Read-only Trello connector for Claude (claude.ai/design)
A tiny remote MCP server that lets Claude on the web read your Trello boards, lists, and cards. There is no official Trello connector, so this fills the gap. It is read-only — it cannot change anything in Trello.
Once deployed you add it under Settings → Connectors → Add custom connector in Claude, and then Claude (including claude.ai/design) can pull your cards.
Tools it exposes
| Tool | What it does |
|---|---|
list_boards |
Your open boards (id, name, URL) |
list_lists |
Lists/columns on a board |
list_cards |
Cards on a board, or in one list |
get_card |
Full card detail: description, labels, due, members, checklists, comments, attachment links |
search_cards |
Find cards by keyword (whole account or one board) |
Step 1 — Get your Trello credentials (read-only)
-
API key: go to https://trello.com/app-key while logged in and copy the Key. (You may be asked to create a Power-Up first — name it anything, e.g. "Claude reader"; that's just how Trello issues a key now.)
-
Read-only token: open this URL in your browser, replacing
YOUR_KEY, then click Allow:https://trello.com/1/authorize?expiration=never&name=ClaudeTrello&scope=read&response_type=token&key=YOUR_KEYscope=readmeans the token can only read — even if the URL leaked, no one could edit your boards with it. Copy the token Trello shows you.
Keep the Key and Token handy for Step 3.
Step 2 — Deploy the server (no local install needed)
The simplest path uses Render's free tier and builds everything in the cloud, so you don't need Node or a newer Python on your Mac.
- Put this folder on GitHub (a private repo is fine). If you have the
ghCLI:
(Or create a repo in the GitHub web UI and upload these files.)git init && git add -A && git commit -m "Trello MCP connector" gh repo create trello-mcp --private --source=. --push - Go to https://render.com → New + → Blueprint → pick the repo. Render
reads
render.yamland sets up a Python 3.11 web service automatically. - When prompted (or under the service's Environment tab) set:
TRELLO_API_KEY→ your key from Step 1TRELLO_TOKEN→ your read-only token from Step 1MCP_PATH→ an unguessable path, e.g./mcp-7f3a9c2e1b(make up your own)
- Deploy. Render gives you a URL like
https://trello-mcp-xxxx.onrender.com.
Your MCP endpoint is that URL plus your MCP_PATH, e.g.
https://trello-mcp-xxxx.onrender.com/mcp-7f3a9c2e1b.
Heads-up on Render free tier: the service sleeps after inactivity, so the first request after a pause takes ~30–60s to wake. Fine for occasional use.
Prefer containers? A
Dockerfileis included — deploy it to Fly.io, Railway, Cloud Run, or your own VPS instead. Same env vars apply.
Step 3 — Add it to Claude
- In Claude (web): Settings → Connectors → Add custom connector.
- Paste your full MCP endpoint URL (URL +
MCP_PATHfrom Step 2). - Save. Claude will connect and discover the five tools above.
Now in claude.ai/design you can say things like "list my Trello boards", "pull the cards from the 'Launch' list on my Marketing board", or "search my Trello for cards mentioning rebrand" — and use them in your design work.
Security — read this
In this simple setup, the URL is the only thing protecting your data: anyone
who has the full URL (including MCP_PATH) can read your Trello. That's why:
MCP_PATHmust be long and random — treat the full URL like a password.- The Trello token is
scope=read, so a leak can't modify your boards. - Don't paste the URL into shared docs, screenshots, or chats.
Want stronger protection? The proper hardening is OAuth on the connector (FastMCP supports it), so Claude authenticates instead of relying on a secret URL. That's a worthwhile upgrade if this holds sensitive boards — say the word and I'll wire it in.
Running locally (optional)
Requires Python 3.10+ (your system Python is 3.9, so use a newer one, e.g. via
pyenv or brew install python@3.11):
python3.11 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env # fill in your key/token
set -a && source .env && set +a
python trello_mcp_server.py
The endpoint will be at http://localhost:8000$MCP_PATH. Local HTTP can't be
added to claude.ai (it needs public HTTPS) — local runs are just for testing,
e.g. with the MCP Inspector (npx @modelcontextprotocol/inspector).
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。