bridge

mcp
Guvenlik Denetimi
Basarisiz
Health Uyari
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 6 GitHub stars
Code Basarisiz
  • child_process — Shell command execution capability in bridge.js
  • execSync — Synchronous shell command execution in bridge.js
  • os.homedir — User home directory access in bridge.js
  • process.env — Environment variable access in bridge.js
  • fs module — File system access in bridge.js
  • network request — Outbound network request in bridge.js
  • child_process — Shell command execution capability in secure-folder-mcp/server.js
  • execSync — Synchronous shell command execution in secure-folder-mcp/server.js
  • exec() — Shell command execution in secure-folder-mcp/server.js
  • process.env — Environment variable access in secure-folder-mcp/server.js
  • fs module — File system access in secure-folder-mcp/server.js
  • network request — Outbound network request in secure-folder-mcp/server.js
  • child_process — Shell command execution capability in secure-folder-mcp/test_client.js
  • fs.rmSync — Destructive file system operation in secure-folder-mcp/test_client.js
  • fs module — File system access in secure-folder-mcp/test_client.js
  • network request — Outbound network request in secure-folder-mcp/test_client.js
  • Hardcoded secret — Potential hardcoded credential in secure-folder-mcp/test_client.js
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

Universal Local-to-Web AI Coding Harness — Connect Local Workspaces to Gemini, Claude, ChatGPT & DeepSeek via Secure MCP

README.md
Bridge Logo

Bridge

Universal Local-to-Web AI Coding Harness

npm version
License: MIT
Node.js
Platform
PRs Welcome

English | 简体中文

Securely connect your local codebases and development environments to Claude Web, ChatGPT, Gemini, and DeepSeek in seconds via the Model Context Protocol (MCP).


Quickstart

1. Instant Execution (Zero Install)

Run directly using npx:

npx @markmayandi/bridge

2. Global Installation

Install globally to access the bridge command anywhere:

npm install -g @markmayandi/bridge

Then navigate to any folder and start:

bridge .

3. Clone and Run from Source

To inspect the source, customize, or contribute:

git clone https://github.com/Manimaran-tech/bridge.git
cd bridge
node bridge.js

Or pass a specific project directory directly:

node bridge.js /path/to/project

Universal MCP Connection

Bridge automatically creates an Anycast HTTPS tunnel and copies the live endpoint to your clipboard:

https://xxxx.trycloudflare.com/mcp

Optional: Fixed MCP URL

By default Bridge uses an ephemeral quick tunnel. If you want a stable URL, create a named Cloudflare tunnel and give Bridge its token:

  1. In Cloudflare Zero Trust → Networks → Tunnels, create a named tunnel and copy its token.
  2. Add a Public Hostname for that tunnel pointing to http://localhost:3000.
  3. Run the one-time setup and paste the token plus your public URL:
bridge setup
# With npx: npx @markmayandi/bridge setup
# Tunnel token: (hidden)
# Public URL: https://mcp.example.com
  1. Start Bridge normally:
bridge .
# With npx: npx @markmayandi/bridge .

The fixed URL is then https://mcp.example.com/mcp, even though the token is stored only for Bridge, in ~/.bridge/config.json (mode 0600). Bridge does not read or modify cloudflared's own ~/.cloudflared configuration. Without a saved token, Bridge keeps using the zero-setup quick tunnel. To return to quick tunnels later, delete ~/.bridge/config.json.

Connecting to Any AI Client

Modern web AI clients (Claude Web, ChatGPT Web, Gemini, and MCP extensions) connect directly using the single /mcp URL.

  1. Copy your live MCP URL from Bridge (automatically placed in your clipboard on startup).
  2. Paste the URL directly into your AI client's MCP or custom tool integration.
  3. Your AI model now has real-time, sandboxed read, write, edit, and terminal test execution capabilities on your local project.

No legacy JSON configuration files or manual port forwarding required.


What Bridge Does

Cloud-based AI models cannot safely access files on your physical machine without open ports or copy-pasting code snippets. Bridge solves this with a high-performance local daemon:

  • Strict Directory Jail: AI operations are strictly restricted to the specified project folder with zero parent-directory traversal.
  • Native Tool Suite: Exposes native coding tools (read, write, edit, str_replace_editor, pwsh, bash, jobs, plan_mode).
  • Zero-Setup Anycast Tunnel: Public HTTPS tunneling powered by Cloudflare with automatic SSL and clipboard sync. Run bridge setup once if you prefer a fixed named-tunnel URL.
  • Terminal Dashboard: Monospaced LN4 Volt interface with live traffic streaming and real-time directory switching.

Security Model

Security Vector Implementation
Authentication & Authorization All HTTP and SSE endpoints enforce Bearer authentication (Authorization: Bearer <token> or ?token=<token>). Unauthenticated requests are rejected with 401 Unauthorized.
Path Traversal Protection Evaluates canonical fs.realpathSync. Access outside the targeted root jail is immediately blocked.
Symlink Boundary Checks Validates symlink resolution. Denies read/write operations targeting locations outside the project folder.
Local Port Isolation Local server binds strictly to 127.0.0.1. Remote traffic routes through secure single-session Anycast tunnels.
Execution Policy & Disablement Shell tools (pwsh, bash, execute_command) are protected by safety policies, restricted to root directory, and can be disabled completely with --disable-shell.
Live Traffic Audit Press [L] at any time to stream every incoming JSON-RPC request and tool call in real time.

Keyboard Controls

Key Action
[C] Re-copies the authenticated MCP URL to your clipboard.
[T] Copies the Bearer authentication token to your clipboard.
[D] Opens the In-Terminal Directory Navigator to switch project folders without restarting.
[L] Toggles real-time streaming traffic and tool execution logs.
[Q] Gracefully closes the tunnel, releases ports, and exits.

Acknowledgements & Credits

Bridge builds upon foundational open-source architecture and local daemon patterns inspired by the DeepSeek harness community. We extend our sincere gratitude to the open-source AI developer ecosystem for advancing local tooling.


License

Distributed under the MIT License. See LICENSE for details.

Maintained by Manimaran

Yorumlar (0)

Sonuc bulunamadi