Summary

This project addresses the chaos of managing scattered AI development tools and MCP servers by providing an enterprise-ready gateway and registry. It offers centralized, secure, and auditable access for both autonomous AI agents and human developers using AI coding assistants. The platform transforms tool sprawl into a governed ecosystem, simplifying configuration, enhancing security, and enabling dynamic tool discovery.

How It Works

The MCP Gateway & Registry acts as a unified access point, abstracting individual MCP servers behind a central gateway. It leverages Nginx for reverse proxying and integrates with identity providers like Keycloak or Cognito for robust OAuth 2.0/3.0 authentication. The system supports dynamic tool discovery, allowing AI agents to find and utilize available tools, and provides a registry for curated server access. It handles multiple transport protocols like SSE and Streamable HTTP, offering a flexible and scalable architecture.

Quick Start & Requirements

• Primary Install: Deploy using pre-built Docker images via ./build_and_run.sh --prebuilt.
• Prerequisites: Docker, git, hf download for sentence-transformers/all-MiniLM-L6-v2 model, environment configuration (.env.example to .env).
• Links: macOS Setup Guide, Pre-built Images Documentation, Quick Start Tutorial, Complete Installation Guide.

Highlighted Details

• MCP Registry CLI: A conversational terminal interface for discovering MCP tools and interacting with the registry.
• Enterprise Security: Integrated vulnerability scanning via Cisco AI Defence MCP Scanner, fine-grained access control, and comprehensive audit trails.
• Anthropic Integration: Seamless import of servers from Anthropic's MCP Registry with full API compatibility.
• Authentication: Supports Machine-to-Machine (M2M) and Three-Legged OAuth (3LO) with Keycloak, Cognito, or other OAuth 2.0 providers.
• Observability: Real-time metrics and monitoring via Grafana, SQLite, and OpenTelemetry.

Maintenance & Community

The project is active with ongoing development on features like ECS Fargate deployment, Microsoft Entra ID authentication, virtual MCP server support, and agent-to-agent protocols. Community engagement is facilitated through GitHub Discussions and Issues.

Licensing & Compatibility

Licensed under the Apache-2.0 License, permitting commercial use and integration with closed-source systems.

Limitations & Caveats

Several key enterprise features, including ECS Fargate deployment, Microsoft Entra ID integration, and advanced virtual server capabilities, are marked as "IN PROGRESS" on the roadmap, indicating they are not yet production-ready. The setup process requires multiple steps and specific dependencies like Docker and an embeddings model.