Live in production

MCP Knowledge Hub

An intelligent gateway, vector search, and unified documentation for your MCP ecosystem

Get started
Documentation
System

System architecture

Modular design with a centralized gateway for maximum scalability and performance

CLIENTE Cursor IDE MCP GATEWAY Node.js + TypeScript search_docs · indexing Qdrant + SQLite (claves) stdio · REST /api VECTOR DB Qdrant Vectors · Search QUEUE Redis Queue · Worker STORAGE PostgreSQL Metadata MCP TOOL search_docs (search over indexed documentation) stdio / JSON-RPC embeddings jobs metadata
Technologies

Tech stack

Built with modern, production-proven technologies

Node.js

Gateway MCP

TypeScript

Typing and MCP SDK

Qdrant

Vector Store

Redis

Cache & Queue

PostgreSQL

Metadata

SQLite

Key index (gateway)

Docker

Containerization

OpenAI

Embeddings

Nginx

Reverse Proxy

Capabilities

Key features

Features designed to maximize your team’s productivity

Vector search

Semantic search powered by OpenAI embeddings and Qdrant to find relevant documentation instantly.

Unified gateway

Single entry point (Node.js + TypeScript) with the MCP `search_docs` tool, stdio, and REST `/api` endpoints for the IDE.

Auto-documentation

Indexing into Qdrant (`mcp_docs`) with folder exclusions, batch embeddings, reindexing on content changes, and deletion sync. Persistent key index in SQLite to avoid full scans. `search_docs` tool for the IDE AI.

Jobs Async

Process heavy tasks in the background using Redis (and optionally a Python worker) without blocking the gateway.

Robust security

Authentication, rate limiting, and operation logging. Per-project MCP configuration (`.cursor/mcp.json`).

Hot Reload

Automatic configuration reload and dynamic discovery of documentation indexed in Qdrant.

Ready to get started?

Set up your MCP gateway in minutes and start building amazing AI-powered experiences.

Download View on GitHub