Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL

mcpjungle/MCPJungle

AI Engineering

Self-hosted gateway that registers MCP servers once and exposes their tools through a single endpoint to AI clients and agents.

1,599
201
Go
MPL-2.0
1 year old

Overview

MCPJungle is a self-hosted MCP gateway for developers and teams managing multiple Model Context Protocol servers. Instead of configuring every MCP server separately in each AI client, you register a server once with MCPJungle and connect Claude, Cursor, Copilot, or custom agents to one MCP endpoint. It centralizes server registration, discovery of tools, prompts and resources, optional tool groups to limit what a client sees, and access-control and observability hooks for shared deployments. The same setup works for a personal local environment or shared team infrastructure.

Quickstart

The documented path starts the gateway container, registers one MCP server, and connects Claude Desktop.

  1. Fetch the compose file and start the server:
curl -O https://raw.githubusercontent.com/mcpjungle/MCPJungle/refs/heads/main/docker-compose.yaml
docker compose up -d

This exposes the streamable HTTP MCP server at http://localhost:8080/mcp by default. A container image is also published on GHCR.

  1. Install the mcpjungle CLI locally, either via Homebrew or from the releases page:
brew install mcpjungle/mcpjungle/mcpjungle
  1. Register a remote MCP server, using Context7 as the example:
mcpjungle register --name context7 --url https://mcp.context7.com/mcp
  1. Add MCPJungle to Claude Desktop's configuration:
{
  "mcpServers": {
    "mcpjungle": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "http://localhost:8080/mcp",
        "--allow-http"
      ]
    }
  }
}
  1. Ask Claude to use the registered server, e.g. request Context7 documentation for /lodash/lodash; it should call the context7__get-library-docs tool through MCPJungle.

Usage Notes

The quickstart requires Docker with Compose to run the gateway, the mcpjungle CLI on the client machine, and a client that supports HTTP MCP connections; the example uses mcp-remote with --allow-http because the local endpoint is plain HTTP. Remote MCP servers registered by URL reach out to those external services, so their availability and policy apply. For current guides, reference material, and operational details, the project points to docs.mcpjungle.com rather than the README, which also hosts a public roadmap; the docs are themselves available as an MCP server.

HotRepo's take

MCPJungle appears once in the HotRepo records provided: on 2026-11-04 it ranked 13th on the Go daily list with 47 stars gained that day, and no all-language daily rank is known. That is a single snapshot, so it supports only that day's placement rather than a trend. The project targets a concrete pain point of MCP adoption — per-client server sprawl — and its registry-plus-gateway design, CLI registration flow, and optional tool grouping are the parts worth evaluating against a team's client and access-control needs.

Project source and documentation

Written by AI from the project's README, updated . The project's own docs are the source of truth.

Hacker News