Skip to main content

MCP Agent Bridge

Bridge AI coding agents as MCP servers for cross-agent code review and collaboration

  • 🔗

    Unified MCP Interface

    Expose Claude, Codex, and Copilot as standard MCP servers. Any MCP client can call any agent using the same protocol.

  • 🔍

    Cross-Agent Code Review

    Get a second opinion from a different AI. Send diffs to Claude for structured review, ask Codex for analysis, or consult Copilot.

  • Singleton Background Services

    Each agent runs as a single background process behind an HTTP proxy. Multiple clients share one instance -- no cold starts.

  • 🖥️

    Cross-Platform

    Standalone binaries for macOS, Linux, and Windows. LaunchAgents, systemd units, or Windows Services -- your choice.

  • 🔒

    Secure by Default

    Read-only tool allowlist, localhost-only binding, input size limits, process timeouts, and working directory validation.

  • 📋

    Structured Output

    Reviews return typed JSON with verdict, issues by severity, and actionable suggestions. Parse results programmatically.