Mcp Bridge— agent skill

<p align="center"> <img src="./docs/assets/mcp-bridge-mark.svg" alt="MCP Bridge" width="120" height="120" /> </p>

<h1 align="center">MCP Bridge</h1>

<p align="center"><strong>Turn internal APIs and cURL workflows into local MCP tools for AI agents.</strong></p>

<p align="center"> MCP Bridge helps teams convert existing HTTP operations into structured, local-first MCP tools that Claude Desktop, Cursor, Zed, and other MCP clients can safely call. </p>

<p align="center"> <a href="./README.md">English</a> · <a href="./README.zh-CN.md">简体中文</a> </p>

<p align="center"> <a href="https://github.com/ainativeera/mcp-bridge/actions/workflows/ci.yml"><img alt="CI" src="https://img.shields.io/github/actions/workflow/status/ainativeera/mcp-bridge/ci.yml?branch=main&label=CI&style=flat-square"></a> <a href="https://github.com/ainativeera/mcp-bridge/blob/main/LICENSE"><img alt="License" src="https://img.shields.io/github/license/ainativeera/mcp-bridge?style=flat-square"></a> <a href="https://github.com/ainativeera/mcp-bridge/stargazers"><img alt="Stars" src="https://img.shields.io/github/stars/ainativeera/mcp-bridge?style=flat-square"></a> <a href="https://github.com/ainativeera/mcp-bridge/issues"><img alt="Issues" src="https://img.shields.io/github/issues/ainativeera/mcp-bridge?style=flat-square"></a> </p>

<p align="center"> <a href="https://github.com/ainativeera/mcp-bridge">GitHub</a> · <a href="#getting-started">Getting Started</a> · <a href="#suggested-mcp-client-config">MCP Config</a> · <a href="./CONTRIBUTING.md">Contributing</a> · <a href="./SECURITY.md">Security</a> </p>

MCP Bridge overview

MCP Bridge is designed for real-world office automation scenarios where teams already have working cURL commands and want to convert them into safer, reusable, and agent-friendly tools without rebuilding integrations from scratch.

Why MCP Bridge

• Import existing cURL commands instead of rebuilding integrations from scratch
• Define MCP-friendly parameter schemas and response descriptions
• Run a local MCP SSE server for desktop AI clients
• Test, debug, and iterate on tools before exposing them to agents
• Package the experience as a desktop app for non-technical users

Core Capabilities

• cURL import: parse common request shapes into editable MCP tools
• Tool editor: configure method, URL, headers, body, parameters, and output fields
• Response shaping: extract the useful parts of large JSON responses
• Local persistence: store tools and logs in SQLite
• MCP transport: expose tools through a local SSE endpoint
• Desktop packaging: ship as an Electron app for macOS and Windows

Use Cases

• Turn internal approval or CRM APIs into agent-callable tools
• Wrap repetitive office workflows behind stable MCP actions
• Prototype agent integrations against existing APIs quickly
• Give AI assistants safe, structured access to business operations

Screenshots

cURL to MCP workflow

cURL import workflow

Local MCP configuration

MCP config panel

Architecture

React UI
  -> Electron desktop shell
  -> Local Express server
  -> SQLite storage
  -> MCP SSE endpoint
  -> External APIs imported from cURL definitions

Key entry points:

• src/App.tsx: main application UI
• server.ts: local Express server and MCP SSE implementation
• electron-main.ts: Electron main process and desktop bridge
• src/db.ts: SQLite persistence layer

Getting Started

Prerequisites

• Node.js 20 or newer
• npm 10 or newer recommended
• macOS for mac desktop packaging
• Windows for native Windows validation, or macOS with Electron Builder cross-packaging support

Install

npm install

Run The Web App

npm run dev

Open:

http://localhost:3000

Run The Desktop App In Development

npm run electron:dev

Build

npm run build

Create Desktop Packages

npm run dist:mac
npm run dist:win

Quality Checks

npm run check

Environment Variables

Copy .env.example to .env and adjust as needed.

Important values:

• PORT: local HTTP and SSE port
• NODE_ENV: development or production
• DB_PATH: SQLite database location
• API_KEY: optional API key for protecting local MCP access
• RATE_LIMIT_WINDOW_MS: API rate limiting window
• RATE_LIMIT_MAX_REQUESTS: max requests allowed in the window
• CORS_ORIGINS: comma-separated origins
• LOG_LEVEL: logger verbosity

MCP Endpoint

When the server is running locally, the SSE endpoint is:

http://localhost:3000/sse

In the packaged desktop app, the UI surfaces the detected local network address to make client setup easier across devices on the same LAN.

Suggested MCP Client Config

{
  "mcpServers": {
    "mcp-bridge": {
      "url": "http://localhost:3000/sse"
    }
  }
}

If you enable an API key:

{
  "mcpServers": {
    "mcp-bridge": {
      "url": "http://localhost:3000/sse",
      "headers": {
        "X-API-Key": "your-secret-api-key"
      }
    }
  }
}

Roadmap

• Safer write-action confirmation for high-risk tools
• Better auth templates for API keys, cookies, and OAuth-style flows
• More robust cURL parsing coverage
• Tool publishing workflow with validation and risk levels
• Richer audit logs, replay, and observability
• Reusable connector templates for office SaaS products

Contributing

Contributions are welcome. Start here:

• Read CONTRIBUTING.md
• Review CODE_OF_CONDUCT.md
• Check SECURITY.md for sensitive reports

Community Standards

• Be respectful and constructive
• Prefer small, focused pull requests
• Include reproduction steps for bugs
• Keep office automation and agent safety in mind when proposing features

License

This project is licensed under the MIT License. See LICENSE.