MCPJungle  by mcpjungle

Self-hosted gateway for AI agents to discover and use tools

created 3 months ago
293 stars

Top 90.1% on SourcePulse

GitHubView on GitHub
Project Summary

MCPJungle acts as a self-hosted registry and gateway for Model Context Protocol (MCP) servers, enabling AI agents to discover and utilize tools from various private MCP servers. It's designed for developers building production-grade AI agents and organizations requiring centralized management, security, and privacy for their AI tool-calling infrastructure.

How It Works

MCPJungle employs a client-server architecture. The server component manages a registry of MCP servers, supporting both Streamable HTTP and STDIO transports. It exposes a unified gateway endpoint (/mcp) that AI clients connect to. When a tool is invoked, MCPJungle routes the request to the appropriate registered MCP server. For STDIO servers, it spawns a new subprocess for each tool call, ensuring no memory leaks but limiting stateful connections.

Quick Start & Requirements

  • Installation: brew install mcpjungle/mcpjungle/mcpjungle or docker pull mcpjungle/mcpjungle.
  • Prerequisites: Docker (for server), npx or uvx (for STDIO servers).
  • Setup: Quickstart involves docker-compose up -d to start the server, followed by mcpjungle register to add MCP servers.
  • Docs: https://github.com/mcpjungle/MCPJungle

Highlighted Details

  • Supports both Streamable HTTP and STDIO MCP server transports.
  • Provides a unified gateway for AI agents (e.g., Claude, Cursor) to access registered tools.
  • Offers enterprise features like Access Control (ACLs) and authentication in production mode.
  • Allows global enabling/disabling of registered tools.

Maintenance & Community

  • Actively developed, with a stated goal of becoming production-ready.
  • Community feedback is encouraged via GitHub Discussions and issues.

Licensing & Compatibility

  • The README does not explicitly state a license. This requires clarification for commercial use or closed-source linking.

Limitations & Caveats

MCPJungle is in BETA and does not maintain long-running connections to MCP servers, meaning a new subprocess is started for every STDIO tool call, impacting performance and preventing stateful connections. OAuth support for authentication is also under development.

Health Check
Last commit

3 days ago

Responsiveness

Inactive

Pull Requests (30d)
14
Issues (30d)
16
Star History
211 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.