Skip to main content
MCPJungle is an open-source, self-hosted gateway for Model Context Protocol servers. Instead of configuring every AI client — Claude, Cursor, Copilot, or a custom agent — to connect to each MCP server individually, you register your MCP servers once in MCPJungle and expose all their tools through a single /mcp endpoint.

​
The problem MCPJungle solves

As you add MCP servers to your workflow, the management overhead compounds. Each new server means updating every AI client’s configuration, duplicating credentials, and losing visibility into which tools are being called. When a server moves or changes, you update it in multiple places. MCPJungle centralizes this. You register MCP servers once in the gateway. Every AI client connects to a single endpoint and automatically gets access to all registered tools — no per-client reconfiguration required.

​
How it works

MCPJungle has a client-server architecture:
  • Gateway server — runs as an HTTP service (default port 8080) and acts as an MCP proxy. It maintains a registry of all your MCP servers, proxies tool calls to the appropriate upstream server, and exposes everything through a single streamable HTTP endpoint at /mcp. The server is typically run via Docker Compose.
  • CLI client — the mcpjungle binary you install locally. Use it to register and deregister MCP servers, manage tool groups, configure access control, and inspect what tools are available.
The gateway supports both streamable HTTP and stdio MCP servers. AI clients connect to MCPJungle over streamable HTTP — MCPJungle is the only MCP server they need to know about.

​
Who should use MCPJungle?

  • Individual developers using Claude Desktop or Cursor who want to access multiple MCP servers without maintaining separate configurations in each client.
  • Teams building AI agents that need centralized access control, observability, and a stable endpoint for tool-calling in production.
  • Organizations that want to manage all MCP client-server interactions from a single, self-hosted location without sending data to third-party services.

​
Get started

Quickstart

Start the gateway, register your first MCP server, and connect Claude in under 5 minutes.

Installation

Install the MCPJungle binary via Homebrew, Docker, or a pre-built binary.

Deploy with Docker

Run the MCPJungle server locally or in production using Docker Compose.

Register MCP servers

Add HTTP and STDIO-based MCP servers to the gateway using the CLI.