Integration Overview

MCPHub is designed to be dropped into an existing MCP setup with minimal friction. The core idea is proxy-based interception: MCPHub sits in front of your MCP servers and transparently forwards requests, logging every tool invocation as it passes through.

What changes

Your MCP clients (Claude Desktop, custom agents, CI scripts) point at MCPHub's proxy URL instead of your server directly.

# Before
https://your-mcp-server.com/mcp

# After
https://mcphub.aniruddha.fyi/api/v1/proxy/{server_id}/mcp

What doesn't change

• Your MCP servers run exactly as before — no changes required
• The MCP protocol is passed through unchanged
• Existing tool schemas, capabilities, and responses are unaffected
• Auth between MCPHub and your server is handled transparently by MCPHub

How the proxy works

MCP Client → MCPHub Proxy → Your MCP Server
               ↓
          Logs tool call
          Publishes to WebSocket
          Triggers alert check

1. Client sends a JSON-RPC request to /api/v1/proxy/{server_id}/mcp
2. MCPHub authenticates the client (JWT or API key)
3. MCPHub fetches the server's auth credentials and forwards the request
4. The response is proxied back to the client unchanged
5. The tool call (name, duration, output size, caller) is logged asynchronously

The total added latency is typically under 5ms.

Integration guides

• Proxy setup — step-by-step for Claude Desktop, custom agents, and scripts
• Per-server auth — configure auth for servers that require it
• Multiple products — managing servers across teams with workspaces
• API key access — programmatic access for CI/CD and scripts
• Direct ingestion — log tool calls without using the proxy