Repolode
Submit
IBM

IBM/mcp-context-forge

An open source AI gateway and registry that federates MCP, A2A, and REST/gRPC APIs behind one endpoint with governance, discovery, and observability.

View on GitHubWebsite
Stars
4,664
Forks
929
Open issues
953
Last commit
9 hours ago

What is it?

ContextForge sits in front of tools, agents, and APIs so AI clients can reach them through a single endpoint. It registers existing MCP servers, A2A agents, and REST or gRPC services, then exposes them as MCP tools and servers with centralized discovery, authentication, rate limiting, retries, and reverse proxying.

The project covers three gateway roles: a tools gateway for MCP and gRPC-to-MCP translation, an agent gateway for A2A and OpenAI- or Anthropic-compatible routing, and an API gateway for REST services. It also includes an Admin UI for managing gateways, tools, and logs, plus OpenTelemetry tracing that works with Phoenix, Jaeger, Zipkin, and other OTLP backends.

It can be installed from PyPI or run as a Docker or Podman container, with PostgreSQL and Redis options for larger deployments. Helm charts and cloud deployment guides are provided for Kubernetes environments.

Key features

  • 01Unified endpoint federating MCP servers, A2A agents, and REST/gRPC APIs
  • 02gRPC-to-MCP translation using server reflection for service discovery
  • 03REST-to-MCP tool adapter with JSON Schema extraction and retry/rate-limit policies
  • 04Admin UI with real-time log viewing and management
  • 05OpenTelemetry tracing with Phoenix, Jaeger, Zipkin, and other OTLP backends
  • 06Plugin system for additional transports and integrations

Use cases

  • Exposing internal REST services to AI agents as MCP tools
  • Giving a team one governed registry of tools for multiple AI clients
  • Adding auth, rate limits, and tracing in front of MCP servers
  • Running a multi-replica gateway on Kubernetes with Redis and PostgreSQL

Quick start

  1. 1

    Create a virtual environment and install from PyPI

    pip install mcp-contextforge-gateway
  2. 2

    Generate secrets into .env.secrets

    python3 -m mcpgateway.scripts.init_secrets
  3. 3

    Start the gateway on port 4444

    mcpgateway --host 0.0.0.0 --port 4444

Things to know

  • Requires Python 3.11 or newer and secrets (JWT_SECRET_KEY, AUTH_ENCRYPTION_SECRET) to be set before the gateway starts.
  • Broad feature scope (40+ plugins, multiple gateway types) means a significant setup and learning curve.

Topics

#mcp#ai-gateway#llm-agents#api-gateway#proxy#observability#kubernetes#fastapi#agents#ai

FAQ

What is mcp-context-forge?
An open source AI gateway and registry that federates MCP, A2A, and REST/gRPC APIs behind one endpoint with governance, discovery, and observability. ContextForge sits in front of tools, agents, and APIs so AI clients can reach them through a single endpoint. It registers existing MCP servers, A2A agents, and REST or gRPC services, then exposes them as MCP tools and servers with centralized discovery, authentication, rate limiting, retries, and reverse proxying.
Is mcp-context-forge free to use? What license does it have?
mcp-context-forge is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is mcp-context-forge still maintained?
Actively maintained. The last commit to mcp-context-forge was on Oct 20, 2026.
What can I use mcp-context-forge for?
Exposing internal REST services to AI agents as MCP tools Giving a team one governed registry of tools for multiple AI clients Adding auth, rate limits, and tracing in front of MCP servers Running a multi-replica gateway on Kubernetes with Redis and PostgreSQL
How do I install mcp-context-forge?
Create a virtual environment and install from PyPI: pip install mcp-contextforge-gateway Generate secrets into .env.secrets: python3 -m mcpgateway.scripts.init_secrets Start the gateway on port 4444: mcpgateway --host 0.0.0.0 --port 4444
Where can I get mcp-context-forge?
The source code and installation instructions are on GitHub: https://github.com/IBM/mcp-context-forge

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects