MCP Context Forge
A self-hosted AI gateway connects MCP servers, agents and REST/gRPC APIs through one endpoint, with access controls and OpenTelemetry tracing.
MCP Context Forge is a gateway and registry for teams connecting AI clients to multiple tools, agents and services. It gives them one place to discover available services, control access and inspect calls.
- Tool and API connections: Combine MCP servers behind one endpoint and expose existing REST or gRPC services as MCP tools, with automatic discovery for gRPC services.
- Agent routing: Connect external agents through A2A, with support for OpenAI-compatible, Anthropic and custom agents.
- Access controls: Apply authentication, user-scoped OAuth tokens, rate limits and retry policies across connected services.
- Shared registries: Manage tools, prompts and resources centrally, including prompt versioning and rollback, resource caching and tool input validation.
- Monitoring and management: Use the browser admin interface to manage services and search or export live logs, and trace calls through OpenTelemetry backends such as Phoenix, Jaeger and Zipkin.
The project is open source under Apache 2.0. You can run the Python gateway locally, including on Windows, or self-host it with Docker. Kubernetes deployments support federation across clusters and Redis-backed caching. Plugins extend its protocols and integrations.
The admin interface supports air-gapped deployment with bundled vendor libraries. Connections to external agents and remote APIs still send requests to those services.
