@letretro/mcp
Enables AI coding tools to access LetRetro retrospectives, boards, cards, and team data through MCP.
README
<div align="center"> <img src="./public/logo.svg" alt="LetRetro" width="120" /> <br/> <br/> <h1>@letretro/mcp</h1> <p><strong>LetRetro MCP — Connect your AI coding tools to LetRetro retrospectives.</strong></p>
<p> <a href="https://www.npmjs.com/package/@letretro/mcp"><img src="https://img.shields.io/npm/v/@letretro/mcp?style=flat&logo=npm&color=%23000" alt="npm version" /></a> <a href="./LICENSE"><img src="https://img.shields.io/badge/license-MIT-%23000?style=flat" alt="MIT License" /></a> <a href="https://nodejs.org"><img src="https://img.shields.io/badge/node-%3E%3D18-%23000?style=flat&logo=nodedotjs" alt="Node >=18" /></a> </p> </div>
What is this?
@letretro/mcp is the official Model Context Protocol package for LetRetro. It gives AI coding assistants direct access to your LetRetro workspace — boards, cards, retrospectives, and team data.
It runs in two modes:
| Mode | What it does |
|---|---|
install |
Interactive CLI wizard that auto-configures your MCP clients |
| server (default) | Local MCP stdio server using the official MCP SDK — forwards requests to mcp.letretro.com |
One install command, zero manual JSON editing.
Quick start
npx @letretro/mcp install
You'll be prompted for your LetRetro API key. The installer detects your installed MCP clients and writes the configuration — you pick which ones to set up.
╔══════════════════════════════════╗
║ LetRetro MCP Installer ║
╚══════════════════════════════════╝
Enter your LetRetro API Key: ••••••
Detected clients:
✓ Claude Desktop
✓ Cursor
✓ VS Code
✓ OpenAI Codex CLI
✓ OpenCode
✓ Antigravity
✓ Hermes
✓ OpenClaw
(Use space to select, enter to confirm)
✔ Select clients to configure:
✓ Claude Desktop: Updated ~/Library/Application Support/Claude/claude_desktop_config.json
✓ Cursor: Created ~/.cursor/mcp.json
╔══════════════════════════════════╗
║ LetRetro MCP Installed ║
╚══════════════════════════════════╝
Restart your MCP clients to start using
LetRetro tools.
What gets configured
The installer adds a letretro server entry to your client's MCP config:
{
"mcpServers": {
"letretro": {
"command": "npx",
"args": ["-y", "@letretro/mcp"],
"env": {
"LETRETRO_API_KEY": "lr_..."
}
}
}
}
Restart your AI coding tool and LetRetro tools are automatically available.
Supported clients
The installer auto-detects these clients and writes the correct config for each:
| Client | Auto-detect | Config file written | Format |
|---|---|---|---|
| Claude Desktop | ✅ macOS/Linux | claude_desktop_config.json |
JSON |
| Cursor | ✅ | ~/.cursor/mcp.json |
JSON |
| VS Code | ✅ binary + config | .vscode/mcp.json |
JSON |
| OpenAI Codex CLI | ✅ | ~/.codex/config.toml |
TOML |
| OpenCode | ✅ | ~/.config/opencode/opencode.json |
JSON |
| Antigravity | ✅ | ~/.gemini/config/mcp_config.json |
JSON |
| Hermes | ✅ | ~/.hermes/config.yaml |
YAML |
| OpenClaw | ✅ | ~/.openclaw/openclaw.json |
JSON |
| Windsurf | ✅ config check | ~/.codeium/windsurf/mcp_config.json |
JSON |
| Continue | ✅ | ~/.continue/config.json |
JSON |
Don't see your client? Open an issue.
How it works
AI Assistant npx @letretro/mcp mcp.letretro.com
│ │ │
│ MCP JSON-RPC over stdio │ │
│ (initialize, tools/list, │ │
│ tools/call, ...) │ │
│──────────────────────────> │ │
│ │ HTTP POST (Bearer token) │
│ │ (forwarded JSON-RPC) │
│ │ ────────────────────────────>│
│ │ │
│ │ JSON-RPC response │
│ │ <────────────────────────────│
│ MCP JSON-RPC response │ │
│ <──────────────────────────│ │
The local process is a proper MCP stdio server built with the official MCP SDK. It:
- Handles the full MCP handshake (
initialize,ping,notifications/*) - Implements
tools/listwith a cached response from the remote server - Forwards
tools/callrequests tomcp.letretro.comand returns results - Supports graceful shutdown via
SIGTERM/SIGINT - Reads
LETRETRO_API_KEYand an optionalLETRETRO_MCP_URLoverride from the environment
All MCP tools, resources, and prompts are implemented server-side on mcp.letretro.com, so you get updates instantly — no package upgrade needed.
Manual setup
If you prefer to configure clients manually instead of using the installer, add the following to your client's MCP config file.
Claude Desktop / Cursor / Antigravity / OpenClaw / Windsurf / Continue:
{
"mcpServers": {
"letretro": {
"command": "npx",
"args": ["-y", "@letretro/mcp"],
"env": {
"LETRETRO_API_KEY": "your_key_here"
}
}
}
}
VS Code:
{
"servers": {
"letretro": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@letretro/mcp"],
"env": {
"LETRETRO_API_KEY": "your_key_here"
}
}
}
}
OpenCode:
{
"mcp": {
"letretro": {
"type": "local",
"command": ["npx", "-y", "@letretro/mcp"],
"enabled": true,
"environment": {
"LETRETRO_API_KEY": "your_key_here"
}
}
}
}
OpenAI Codex CLI:
[mcp_servers.letretro]
command = "npx"
args = ["-y", "@letretro/mcp"]
[mcp_servers.letretro.env]
LETRETRO_API_KEY = "your_key_here"
Hermes:
mcp_servers:
letretro:
command: "npx"
args:
- "-y"
- "@letretro/mcp"
env:
LETRETRO_API_KEY: "your_key_here"
Environment variables
| Variable | Default | Description |
|---|---|---|
LETRETRO_API_KEY |
— | Your LetRetro API key (required) |
LETRETRO_MCP_URL |
https://mcp.letretro.com |
Override the remote MCP server URL (for testing) |
LETRETRO_VALIDATE_URL |
https://mcp.letretro.com |
URL used for API key validation during install |
Development
git clone https://github.com/LetRetroHQ/mcp-cli.git
cd letretro-mcp/mcp-npm
pnpm install
pnpm dev # watch mode
pnpm build # production build
pnpm typecheck # type checking
pnpm test # run tests
Project structure
src/
├── index.ts # Entry point — routes to install or server mode
├── install.ts # Interactive installer (prompt → detect → configure)
├── server.ts # MCP stdio server (MCP SDK) — forwards to mcp.letretro.com
├── clients/
│ ├── index.ts # Client registry and detection helpers
│ ├── claude.ts # Claude Desktop
│ ├── cursor.ts # Cursor
│ ├── vscode.ts # VS Code
│ ├── codex.ts # OpenAI Codex CLI (TOML config)
│ ├── opencode.ts # OpenCode
│ ├── antigravity.ts # Google Antigravity
│ ├── hermes.ts # Hermes Agent (YAML config)
│ ├── openclaw.ts # OpenClaw
│ ├── windsurf.ts # Windsurf
│ └── continue.ts # Continue
└── utils/
├── paths.ts # OS-aware config file paths
├── toml.ts # TOML serializer for Codex config
└── yaml.ts # YAML serializer for Hermes config
Contributing
Contributions are welcome! Whether it's adding support for a new MCP client, improving the installer, or fixing a bug:
- Fork the repo
- Create a feature branch (
git checkout -b feat/my-feature) - Commit your changes (
git commit -am 'feat: add support for...') - Push to the branch (
git push origin feat/my-feature) - Open a Pull Request
See CONTRIBUTING.md for detailed guidelines.
License
<div align="center"> <sub>Built by the LetRetro team</sub> </div>
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。