Hatago MCP Hub

Hatago MCP Hub is a lightweight relay that connects various MCP servers to your AI tools, enabling seamless access to multiple model-context providers from tools like Claude Code, Codex CLI, Cursor, Windsurf, and VS Code. It simplifies starting, routing, and managing MCP servers from a single gateway.

How to use

You connect your MCP client tools to Hatago MCP Hub, then start the hub in either STDIO mode (local, script-driven integration) or HTTP mode (remote, network access). The hub aggregates MCP servers, forwards requests, and provides progress updates and notifications to your tools. You can filter which servers to run by tags, observe live progress, and use environment variable expansion for dynamic configurations.

How to install

Prerequisites: Node.js and npm or a compatible runtime environment. You will also need a compatible MCP client or CLI to connect through Hatago.

# Quick start without installation
npx @himorishige/hatago-mcp-hub init

# Start in STDIO mode (for Claude Code)
# NOTE: STDIO requires a config file path
npx @himorishige/hatago-mcp-hub serve --stdio --config ./hatago.config.json

# Or start in HTTP mode without a config (demo/dev)
npx @himorishige/hatago-mcp-hub serve --http
```}]} ,{

Optional: you can install Hatago MCP Hub globally or add it as a project dependency for repeated usage.

Configuration and connectivity

Hatago supports both local STDIO servers and remote HTTP-based MCP endpoints. You can start the hub with a preconfigured hatago.config.json, connect clients via STDIO, or connect through HTTP to a running hub.

Security and runtime notes

If you enable metrics, you can expose lightweight in-memory metrics over HTTP at /metrics. You can also control logging behavior with environment variables like HATAGO_METRICS and HATAGO_LOG.

Troubleshooting

Common issues include transport or server connection failures. Use verbose logging to diagnose problems and verify that environment variables used by remote MCP servers are correctly set.

Performance considerations

Hatago emphasizes fast startup and lightweight operation, while providing straightforward configuration options and direct server management.

Examples of typical usage scenarios

• Connect Claude Code or Gemini CLI to Hatago in STDIO mode and point to a hatago.config.json that defines your MCP servers.

Available tools

startServer

Programmatic API to start the MCP hub with configuration options.

createHub

Create a lightweight hub instance for programmatic usage in your application.

listTools

List available tools connected to the hub.

hub events

Event system for metrics and logs, e.g., 'tool:called'.