MCP Gateway & Registry
A self-hosted MCP gateway and AI asset registry centralizes discovery, OAuth access and audit logs for MCP servers, agents and skills.

MCP Gateway & Registry gives platform teams a shared catalog and access gateway for the tools their AI agents and coding assistants use. It addresses scattered server connections, credentials and access policies across an organization.
- Asset discovery: Search MCP servers, A2A agents, versioned skills and custom asset types by natural language or keywords, with federation across Anthropic's MCP Registry, AWS Agent Registry and peer registries.
- Shared tool endpoints: Combine tools from multiple backends into virtual MCP servers with access controls for individual tools.
- Identity and credentials: Connect Keycloak, Entra ID, Okta and other supported identity providers, and broker per-user credentials for SaaS MCP servers such as Slack, Atlassian and GitHub without storing those tokens on users' laptops.
- Security and auditing: Scan registered servers, agents and skills, hold unsafe items for review, and record attributable access and administration events with credentials masked.
- Inference and REST gateway: Optionally route requests to OpenAI, Amazon Bedrock, OpenAI-compatible inference endpoints and generic REST services under shared access policies and audit logging.
The project is open source under Apache 2.0. It runs through Docker Compose or Kubernetes, including Amazon EC2 and EKS deployments, or on Amazon ECS. Registry data lives in MongoDB or DocumentDB. A2A agents communicate directly by default after registry discovery and authorization; their traffic doesn't all pass through the gateway.
Connections to external SaaS and cloud inference services send requests to those providers. Anonymous usage telemetry is enabled by default and can be disabled; it excludes credentials, endpoints, model names and personal information.