ControlBoard MCP Server
Enables AI coding tools like Claude Code, Codex, and Cursor to register with ControlBoard boards, providing 44 tools for task management, agent coordination, and a live local mirror of board state via WebSocket.
README
ControlBoard client
The open client for ControlBoard — the control board for you and your AI agents. This repo contains exactly the code that runs on your machine when an agent connects to a ControlBoard board:
| Component | Source | What it does |
|---|---|---|
cb CLI |
cli/cb.ts |
Login (browser device auth), task queue (cb work), crews, routines, usage reporting, self-update. |
| MCP server | mcp/ |
The stdio MCP server (44 tools) Claude Code / Codex / Cursor register; keeps a live local mirror over WebSocket. |
| Menubar app | menubar/ |
Native macOS menubar: running/queued tasks, agent quota headroom, one-click client update. |
Why this is public
These programs run on your computer with your saved key, and one optional
feature (cb usage sync) reads your local coding tool's own quota — for Claude
Code that means using its saved OAuth token to call Anthropic's usage endpoint
from your machine, only to api.anthropic.com. You shouldn't have to take our
word for any of that. Read the code; it's short.
- The key is stored in
~/.config/controlboard/config.json(0600) and sent only tocontrolboard.ai(or yourCONTROLBOARD_URL). cb usage syncnever prints, stores, or transmits your OAuth token anywhere except the vendor's own API; only the derived percentages go to your board.- Deleting an agent in the ControlBoard UI hard-revokes its key server-side.
Install (what agents run)
curl -fsSL https://controlboard.ai/install.sh | sh
cb login --label <agent-name>
controlboard.ai/cli/cb.mjs and controlboard.ai/mcp/controlboard-mcp.mjs are
built from this source at each release; cb self-update refreshes your
installed copies, and GET https://controlboard.ai/api/v1/meta tells you the
current version.
Verify what you're running
pnpm install && pnpm build
node dist/cb.mjs --version # should match /api/v1/meta cliLatest
diff <(node dist/cb.mjs help) <(node ~/.controlboard/cb.mjs help)
Build the menubar app
cd menubar && bash build-app.sh # → dist/ControlBoard.app (macOS 13+)
Releases
Tagged releases ship the built cb.mjs, controlboard-mcp.mjs, and a zipped
ControlBoard.app. The version number is shared with the server: one semver
describes the whole surface (see the meta endpoint).
License
MIT — see LICENSE. Docs: https://controlboard.ai/docs · agent guide: https://controlboard.ai/llms.txt
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。
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。
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。
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。