MCPJungle

MCPJungle acts as a centralized gateway and registry for MCP servers and their tools. It lets you register remote MCP servers and expose their tools through a single endpoint, enabling AI agents to discover and call those tools via the MCP protocol. The gateway runs over a streamable HTTP transport and serves tool calls at the /mcp endpoint, providing a unified interface for clients like Claude, Cursor, or Copilot to access all registered MCP servers.

To use MCPJungle, first start the gateway (preferably in Docker or via a binary). Then install the MCPJungle CLI on your client machine and register any MCP servers you want to expose. Once registered, you can configure your MCP clients to route tool calls through MCPJungle. For example, you can point Claude to MCPJungle as an MCP registry and have it call a registered server’s tool using the provided mcpServers entry. The README demonstrates a typical workflow: start MCPJungle, register a remote server (e.g., context7), and configure Claude to access the tools via MCPJungle. Tools in MCPJungle can include streamable HTTP-based servers or stdio-based servers, depending on how you set up the underlying MCP servers.

After registration, you can manage and discover all registered MCP servers and their tools from a single place, enabling easier access control and observability for your AI agents.

Prerequisites:

• Docker and Docker Compose (recommended for local run)
• Optional: Homebrew (macOS) for installing the MCPJungle binary
• Network access to pull images and download release assets

1. Quick start via Docker (recommended for local testing)

• Pull and run MCPJungle with Docker: curl -O https://raw.githubusercontent.com/mcpjungle/MCPJungle/refs/heads/main/docker-compose.yaml docker compose up -d

• Verify the gateway is running: curl http://localhost:8080/health

• If you plan to register stdio-based MCP servers that rely on npx or uvx, use the stdio-enabled image tag: MCPJUNGLE_IMAGE_TAG=latest-stdio docker compose up -d

2. Alternative: run via binary (Mac/Linux)

• Install via Homebrew (macOS): brew install mcpjungle/mcpjungle/mcpjungle

• Verify installation: mcpjungle version

• If not using Homebrew, you can download the binary from Releases and place it in your PATH, then run: mcpjungle version

3. Alternative: standalone Docker run (gateway only)

• Run the gateway in Docker with a minimal image: docker run -d --name mcpjungle -p 8080:8080 ghcr.io/mcpjungle/mcpjungle

4. Register MCP servers (client-side step)

• Install the MCPJungle CLI on your local machine (brew or release page): brew install mcpjungle/mcpjungle/mcpjungle

• Register a remote MCP server (example using context7): mcpjungle register --name context7 --url https://mcp.context7.com/mcp

5. Configure a client (example for Claude)

• In your Claude MCP configuration, point to MCPJungle as the MCP gateway using a setup like the following (already demonstrated in the README): { "mcpServers": { "mcpjungle": { "command": "npx", "args": [ "mcp-remote", "http://localhost:8080/mcp", "--allow-http" ] } } }

Notes and tips:

• MCPJungle supports both standard image deployments and a dedicated stdio image for servers that rely on npx or uvx. Use the stdio tag if you rely heavily on stdio-based MCP servers like filesystem, time, github, etc.
• When deploying in production, consider using a separate Postgres database and configure the DATABASE_URL in MCPJungle to point to it for persistence.
• The gateway endpoint is /mcp and uses a streamable HTTP transport for efficient tool calls.
• The enterprise features (Access Control, OpenTelemetry) can be enabled via configuration in production deployments; consult the repository docs for specifics.
• If you update the docker-compose.yaml in your environment, you may need to adjust environment-specific settings like enterprise mode or database connections.
• Tools and servers registered in MCPJungle can be managed via the CLI; use mcpjungle register, deregister, and other CLI commands as described in the Quickstart guide.