MCP Gateway Registry

by agentic-community

MCP Gateway Registry is an enterprise-grade control plane for managing MCP servers, AI agents, and agent-to-agent communication from a single dashboard. Instead of configuring each MCP server individually in Claude, Cursor, or Codex, you route everything through one gateway that handles authentication, discovery, version routing, and audit logging. Built by the agentic-community organization, it has 562 GitHub stars, 126 forks, and 903 commits — indicating active, sustained development rather than a weekend project. The standout feature is agent-to-agent (A2A) communication. Agents can register themselves, discover other agents via semantic search, and communicate directly through the A2A protocol. This turns isolated MCP tools into a collaborative mesh where your code review agent can talk to your deployment agent without routing through a human. Enterprise security is not an afterthought. The gateway supports OAuth 2LO/3LO flows, multi-provider IAM (Keycloak, Microsoft Entra ID, Okta, Auth0), fine-grained scope-based access control, and M2M service accounts with self-signed JWT tokens. Credential masking in audit logs, integrated vulnerability scanning via Cisco AI Defense scanners, and PKI-based trust verification through the Agent Name Service (ANS) round out the security stack. Version routing with instant rollback means you can deploy a new MCP server version, test it with a subset of agents, and roll back in seconds if it breaks. Registry federation lets multiple instances sync across teams or organizations — useful for enterprises with separate dev, staging, and production environments. Virtual MCP server aggregation combines tools from multiple backend servers into a single virtual endpoint. Your AI agent sees one server with 50 tools instead of 10 servers with 5 tools each. Deploy via Docker Compose for local development, Kubernetes for production, or AWS ECS/Fargate with Terraform for multi-AZ cloud deployments. Observability ships with Prometheus metrics and OTLP direct push export. Apache 2.0 licensed. Backend in Python, frontend in JavaScript/React. For a simpler MCP aggregation approach, compare with MetaMCP. For the AI agents that consume these tools, explore Tech Leads Club Agent Skills for verified skill packages.

FAQ