MCPHub

MCPHub is an MCP server that acts as a self-hosted gateway and proxy to connect, manage, and scale multiple Model Context Protocol servers. It connects to local and remote MCP servers across stdio, Streamable HTTP, and SSE transports, as well as external databases like PostgreSQL when running in database mode. Software developers, AI engineers, and platform administrators use MCPHub to centralize their MCP toolsets behind unified endpoints for AI clients such as Claude Desktop, Claude Code, Cursor, Cherry Studio, and OpenWebUI. The platform enables users to organize tools into specific groups, enforce access control through OAuth 2.0 and bearer keys, execute semantic tool discovery via vector smart routing, and monitor tool-call activity through a browser dashboard or CLI. By aggregating distributed MCP servers into structured HTTP endpoints, it allows AI coding assistants and LLM applications to access multiple tool ecosystems without individual configuration changes or manual endpoint remapping.

Category: AI & LLM Tooling

Tags: gateway, infrastructure, proxy, sse

Visit MCPHub

How to install and configure MCPHub

1. Prepare an mcp_settings.json file configuring your desired servers (using commands like npx or uvx). 2. Run MCPHub using Docker: docker run -p 3000:3000 -v ./mcp_settings.json:/app/mcp_settings.json -v ./data:/app/data samanhappy/mcphub. 3. Access the web dashboard at http://localhost:3000 using the username admin and either the generated log password or your pre-set ADMIN_PASSWORD. 4. Create a bearer key in the dashboard or terminal to authorize client traffic. 5. Register the hub endpoint in your MCP client (such as Cursor or Claude Desktop) using an SSE or HTTP entry pointing to http://localhost:3000/mcp for all servers, http://localhost:3000/mcp/{group} for a group, or http://localhost:3000/mcp/$smart for smart routing.

What you can do with MCPHub

• Consolidating dozens of local stdio and remote SSE MCP servers into a single stable HTTP endpoint for Cursor or Claude. - Organizing AI tools into access-controlled team groups with bearer keys and OAuth 2.0 authorization rules. - Performing semantic vector search across tool definitions to discover relevant endpoints using smart routing routes. - Monitoring tool latency, request status, connection health, and real-time execution logs from a browser dashboard. - Managing server definitions, listing tools, and invoking commands directly from a terminal using the built-in CLI.

Key facts

• Open Source
• https://github.com/samanhappy/mcphub
• AI & LLM Tooling, Developer Tools & Code Intelligence
• gateway, infrastructure, proxy, sse

Part of MCP Servers

How do I install MCPHub?

You can deploy MCPHub using Docker with the command docker run -p 3000:3000 -v ./data:/app/data samanhappy/mcphub. Alternatively, you can clone the repository from GitHub, run pnpm install, and execute pnpm dev for local development. A preconfigured mcp_settings.json file can be mounted to define servers upon launch.

What can MCPHub do?

MCPHub aggregates multiple MCP servers across stdio and SSE into unified HTTP routes. It features group management, AI-powered semantic smart routing, request logging, tool result compression, health checks, an interactive web dashboard, CLI management, and OAuth 2.0 or bearer key authentication.

Which MCP clients work with MCPHub?

MCPHub exposes standard Model Context Protocol endpoints compatible with any client supporting Streamable HTTP or SSE connections. Supported clients include Claude Desktop, Claude Code, Cursor, Cherry Studio, and OpenWebUI. AI clients can connect to individual server routes, specific group endpoints, or smart routing selectors.

Is MCPHub open source?

Yes, MCPHub is open-source software distributed under the Apache 2.0 License. The source code, documentation, and Docker configuration files are publicly accessible in the samanhappy/mcphub repository on GitHub.