Gateway Registry

---

What is MCP Gateway & Registry?

The MCP Gateway & Registry is a unified platform designed for centralizing access to both MCP Servers and AI Agents using the Model Context Protocol (MCP). It serves three core functions:

1. Unified MCP Server Gateway – Centralized access point for multiple MCP servers
2. MCP Servers Registry – Register, discover, and manage access to MCP servers with unified governance
3. Agent Registry & A2A Communication Hub – Agent registration, discovery, governance, and direct agent-to-agent communication through the A2A (Agent-to-Agent) Protocol

The platform integrates with external registries such as Anthropic's MCP Registry (and more to come), providing a single control plane for both tool access, agent orchestration, and agent-to-agent communication patterns.

Why unified? Instead of managing hundreds of individual MCP server configurations, agent connections, and separate governance systems across your development teams, this platform provides secure, governed access to curated MCP servers and registered agents through a single, unified control plane.

Transform this chaos:

❌ AI agents require separate connections to each MCP server
❌ Each developer configures VS Code, Cursor, Claude Code individually
❌ Developers must install and manage MCP servers locally
❌ No standard authentication flow for enterprise tools
❌ Scattered API keys and credentials across tools
❌ No visibility into what tools teams are using
❌ Security risks from unmanaged tool sprawl
❌ No dynamic tool discovery for autonomous agents
❌ No curated tool catalog for multi-tenant environments
❌ A2A provides agent cards but no way for agents to discover other agents
❌ Maintaining separate MCP server and agent registries is a non-starter for governance
❌ Impossible to maintain unified policies across server and agent access

Into this organized approach:

✅ AI agents connect to one gateway, access multiple MCP servers
✅ Single configuration point for VS Code, Cursor, Claude Code
✅ Central IT manages cloud-hosted MCP infrastructure via streamable HTTP
✅ Developers use standard OAuth 2LO/3LO flows for enterprise MCP servers
✅ Centralized credential management with secure vault integration
✅ Complete visibility and audit trail for all tool usage
✅ Security features with governed tool access
✅ Dynamic tool discovery and invocation for autonomous workflows
✅ Registry provides discoverable, curated MCP servers for multi-tenant use
✅ Agents can discover and communicate with other agents through unified Agent Registry
✅ Single control plane for both MCP servers and agent governance
✅ Unified policies and audit trails for both server and agent access

┌─────────────────────────────────────┐     ┌──────────────────────────────────────────────────────┐
│          BEFORE: Chaos              │     │    AFTER: MCP Gateway & Registry                     │
├─────────────────────────────────────┤     ├──────────────────────────────────────────────────────┤
│                                     │     │                                                      │
│  Developer 1 ──┬──► MCP Server A    │     │  Developer 1 ──┐                  ┌─ MCP Server A    │
│                ├──► MCP Server B    │     │                │                  ├─ MCP Server B    │
│                └──► MCP Server C    │     │  Developer 2 ──┼──► MCP Gateway   │                  │
│                                     │     │                │    & Registry ───┼─ MCP Server C    │
│  Developer 2 ──┬──► MCP Server A    │ ──► │  AI Agent 1 ───┘         │        │                  │
│                ├──► MCP Server D    │     │                          │        ├─ AI Agent 1      │
│                └──► MCP Server E    │     │  AI Agent 2 ──────────────┤        ├─ AI Agent 2     │
│                                     │     │                          │        │                  │
│  AI Agent 1 ───┬──► MCP Server B    │     │  AI Agent 3 ──────────────┘        └─ AI Agent 3     │
│                ├──► MCP Server C    │     │                                                      │
│                └──► MCP Server F    │     │              Single Connection Point                 │
│                                     │     │                                                      │
│  ❌ Multiple connections per user  │     │         ✅ One gateway for all                      │
│  ❌ No centralized control         │     │         ✅ Unified server & agent access            │
│  ❌ Credential sprawl              │     │         ✅ Unified governance & audit trails        │
└─────────────────────────────────────┘     └──────────────────────────────────────────────────────┘

Note on Agent-to-Agent Communication: AI Agents discover other AI Agents through the unified Agent Registry and communicate with them directly (peer-to-peer) without routing through the MCP Gateway. The Registry handles discovery, authentication, and access control, while agents maintain direct connections for efficient, low-latency communication.

Unified Agent & Server Registry

This platform serves as a comprehensive, unified registry supporting:

• ✅ MCP Server Registration & Discovery – Register, discover, and manage access to MCP servers
• ✅ AI Agent Registration & Discovery – Register agents and enable them to discover other agents
• ✅ Agent-to-Agent (A2A) Communication – Direct agent-to-agent communication patterns using the A2A protocol
• ✅ Multi-Protocol Support – Support for various agent communication protocols and patterns
• ✅ Unified Governance – Single policy and access control system for both agents and servers
• ✅ Cross-Protocol Agent Discovery – Agents can discover each other regardless of implementation
• ✅ Integrated External Registries – Connect with Anthropic's MCP Registry and other external sources
• ✅ Agent Cards & Metadata – Rich metadata for agent capabilities, skills, and authentication schemes

Key distinction: Unlike separate point solutions, this unified registry eliminates the need to maintain separate MCP server and agent systems, providing a single control plane for agent orchestration, MCP server access, and agent-to-agent communication.

MCP Servers, Agents and Skills Registry

Watch how MCP Servers, A2A Agents, and External Registries work together for dynamic tool discovery:

https://github.com/user-attachments/assets/97c640db-f78b-4a6c-9662-894f975f66e2

---

MCP Tools in Action

View MCP Tools Demo

---

MCP Registry CLI

Interactive terminal interface for chatting with AI models and discovering MCP tools in natural language. Talk to the registry using a Claude Code-like conversational interface with real-time token status, cost tracking, and AI model selection.

Quick Start: registry --url https://mcpgateway.ddns.net | Full Guide

---

What's New

• MongoDB Atlas and Replica-Set Support via Connection String Override - The storage layer now accepts a full MongoDB connection string via the new MONGODB_CONNECTION_STRING environment variable, enabling MongoDB Atlas (mongodb+srv://...), replica sets, and any URI-level tuning not expressible through the discrete DOCUMENTDB_* variables. When set, the URI takes precedence over DOCUMENTDB_HOST/PORT/USERNAME/PASSWORD/DATABASE; when empty, the registry falls back to the existing variables so all current deployments keep working unchanged. Wired through Docker Compose, Helm/EKS, and the ECS Terraform module (with a Secrets Manager variant preferred for URIs containing credentials to keep them out of Terraform state). In addition to MongoDB CE and Amazon DocumentDB, MongoDB Atlas is now a supported backend. See .env.example and terraform/aws-ecs/terraform.tfvars.example for the new parameter. FAQ: How do I configure MongoDB Atlas instead of MongoDB CE?

• Group-Restricted Agent Visibility - Agent publishers can now restrict which IdP groups can see their agent by setting visibility: "group-restricted" and specifying allowedGroups at registration time, without needing an admin to change IAM scopes. Works as a second filter on top of the existing IAM group scope layer: users must pass both the IAM scope check and the allowed_groups check. Nginx forwards JWT group claims via X-Groups header, the list endpoint enforces group filtering for all non-admin users, and the CLI supports --allowed-groups for both registration and filtering. Frontend registration and edit forms include a Visibility dropdown and Allowed Groups input. Compatible with all supported IdPs (Keycloak, Entra ID, Cognito, Okta, Auth0). (#883, #922) Full Guide | FAQ

• Admin Data Export - Download registry data as JSON files for debugging, auditing, and backup. A new Data Export section in the admin Settings page supports 11 collections: Servers, Agents, Skills, Virtual Servers, Federation Peers, Federation Configs, Registry Card, IAM Users, IAM Groups, IAM M2M Clients, and Scopes. Download individual collections or use the Download All as ZIP button (powered by JSZip) with per-collection progress indicators. Includes a sensitive data warning banner and a dedicated scopes export endpoint that dumps full server_access rules. Admin-only access, not visible to non-admin users.

• Centralized Log Rotation, Storage, and Retrieval - Production-grade application logging with RotatingFileHandler (50 MB, 5 backups) for both the registry and auth-server. Optional MongoDB storage via a non-blocking MongoDBLogHandler with buffered background writes and TTL-based auto-expiry. Admin REST API endpoints (GET /api/admin/logs for querying with filters, GET /api/admin/logs/export for JSONL download, GET /api/admin/logs/metadata for available services and levels) and a Settings UI Log Viewer with filtering by service, level, hostname, search text, and time range. Security includes MongoDB regex injection prevention via re.escape(), rate limiting (10 requests per 60 seconds per user), and max search length validation. MongoDB logging is OFF by default; enable with APP_LOG_MONGODB_ENABLED=true. File-based rotation is always active.

• Registration Webhooks and Gate - Two external integration points for registration lifecycle events. Registration Gate (Admission Control): call an external endpoint to approve or deny registration and update requests before they are persisted. Supports all asset types (servers, agents, skills) for both register and update operations. Fail-closed design: if the gate endpoint is unreachable after configurable retries with exponential backoff, the registration is blocked. Sensitive fields (credentials, tokens, passwords) are automatically stripped from the payload sent to the gate. Supports Bearer token, API key, or unauthenticated access. Gate returns 200 to allow, 403 to deny with a custom error message. Registration Webhooks: send HTTP POST notifications to an external URL when servers, agents, or skills are registered or deleted. Enables real-time integration with CMDBs, CI/CD pipelines, Slack, or any external system. Fire-and-forget delivery (failures are logged, never block the caller). Supports Bearer token and custom API key authentication with configurable headers and timeouts. Both are configured across Docker Compose, Terraform/ECS, and Helm/EKS. (#809, #742) Webhooks and Gate Guide

• Multi-Key Static Tokens with Per-Key Groups - Replace the single REGISTRY_API_TOKEN with a REGISTRY_API_KEYS JSON object where each key carries its own name, secret, and group list. Groups resolve to scopes through the standard group_mappings pipeline, so each key gets exactly the privileges its groups grant. Supports zero-downtime rotation (add new key, migrate clients, remove old key), timing-safe comparison via hmac.compare_digest, and full coexistence with the legacy single token. The legacy token is auto-promoted to a "legacy" entry in the internal token map with mcp-registry-admin scopes for backward compatibility. Key names appear in audit logs as the username for traceability. Scope changes propagate to static tokens via the auth server reload mechanism without requiring a restart. (#779) Registry API Authentication Guide | FAQ

• Registry API Authentication: Unified Model - The Registry API (/api/*, /v0.1/*) accepts four credential types concurrently: session cookies (browser UI), IdP-issued JWTs (Okta/Entra/Cognito/Keycloak) and UI-issued self-signed JWTs, a static REGISTRY_API_TOKEN for trusted service-to-service callers, and a separate FEDERATION_STATIC_TOKEN scoped to federation/peer endpoints only. Previously, enabling static-token mode silently blocked JWT callers on /api/* (#871). One improvement remains on the roadmap: external user access tokens that let a frontend app call the registry on behalf of its logged-in users without sharing a registry credential (#826). Registry API Authentication Guide | FAQ

• GitHub Private Repository Auth for Agent Skills - Fetch SKILL.md files from private GitHub repositories using Personal Access Tokens (PAT) or GitHub App authentication (recommended for organizations). Supports GitHub Enterprise Server with configurable API base URLs and extra host matching. Auth headers are only sent to github.com, raw.githubusercontent.com, and explicitly listed hosts. Helm deployments support Kubernetes secrets for credential injection. Configured across Docker Compose, Helm, and Terraform/ECS. See .env.example for all parameters. Configuration Guide

• Configurable Tab Visibility, Pagination, and Lifecycle Filtering - Four new SHOW_*_TAB environment variables control which dashboard tabs are visible independently of REGISTRY_MODE (formula: tab_visible = mode enables feature AND SHOW_*_TAB). All APIs (GET /api/servers, /api/agents, /api/skills) now support limit/offset pagination. New lifecycle status filtering in the sidebar lets you filter by active, deprecated, or experimental assets. Also adds GET /api/servers/{path} for single server retrieval, improved semantic search ranking (global ranking replaces the old 3-per-type cap), network-trusted JWT token generation, and heartbeat telemetry opt-out. See .env.example for all new configuration parameters. Tab Visibility Configuration

• AWS Agent Registry Federation - Federate MCP servers, A2A agents, and agent skills from AWS Agent Registry into MCP Gateway Registry. Add multiple AgentCore registries (same or different AWS accounts/regions), select which descriptor types to sync (MCP, A2A, CUSTOM, AGENT_SKILLS), and manage everything from the External Registries settings page. Supports cross-account access via IAM role assumption, cascade cleanup on registry removal, and automatic sync on startup. Enable with a single environment variable (AWS_REGISTRY_FEDERATION_ENABLED=true) for ECS/Terraform or Helm deployments. Operations Guide | Design Document

• Register Any Agent (A2A and Non-A2A) - The Agent Registry now supports registering any agent, not just A2A protocol agents. A new supported_protocol field (a2a or other) distinguishes agent types. Register through the UI (checkbox for A2A agents, dropdown for protocol selection on edit), the API (supportedProtocol field in registration payload), or the CLI (--supported-protocol flag). Default trust_level updated to community and visibility to public for consistency. A one-time backfill script normalizes existing agents in MongoDB. Two new Claude Code skills -- generate-agent-card and generate-server-card -- analyze source code and generate registration-ready agent or server card JSON files. Documentation

• Amazon Bedrock AgentCore Bulk Import - Auto-discover and register all AgentCore Gateways and Agent Runtimes from your AWS account in a single command. The CLI scans for READY resources, registers gateways as MCP Servers and runtimes as MCP Servers or A2A Agents based on protocol, and writes a token refresh manifest for automated credential rotation. Supports multi-account scanning, OIDC-compliant identity providers (Cognito, Auth0, Okta, Entra ID, Keycloak), and overwrite mode for updating existing registrations. AgentCore Operations Guide | Design Document

• Anonymous Usage Telemetry - Privacy-first telemetry to track registry adoption patterns. Sends only non-sensitive deployment metadata (version, OS, storage backend, auth provider) -- no PII, no hostnames, no user data. Opt-out by default (startup ping is ON, set MCP_TELEMETRY_DISABLED=1 to disable). Opt-in daily heartbeat with aggregate counts (server/agent/skill totals). HMAC-signed requests, IP-hashed rate limiting, strict schema validation, and fail-silent design ensure zero impact on registry operation. Admin API to force heartbeat/startup events on demand. Telemetry Documentation

• Agent Name Service (ANS) Integration - Adds PKI-based trust verification for registered agents and MCP servers through GoDaddy's Agent Name Service. Agent owners link their ANS Agent ID to their registry entry, and the registry verifies the identity via the ANS API, displaying a clickable trust badge on agent cards and semantic search results. A background scheduler re-verifies all linked identities every 6 hours with circuit breaker protection. Supports verified, expired, and revoked status tracking with admin endpoints for manual sync, metrics, and health checks. Design and Operations Guide | Demo Video

• Registry Card for Federation Discovery - As registries increasingly need to discover and communicate with each other, we've implemented the Registry Card specification—a standardized discovery document accessible via /.well-known/registry-card. This provides essential metadata including authentication endpoints, capabilities, and contact information for any registry instance. Enhanced server, agent, and skills cards with richer metadata enable better federation workflows. Registry Card Configuration Guide

• 🔑 Auth0 Identity Provider Support - Full enterprise SSO integration with Auth0 as an identity provider. The harmonized IAM API now supports Auth0 alongside Keycloak, Microsoft Entra ID, and Okta, providing a unified interface to create users, groups, and M2M service accounts regardless of your IdP choice. Features include Auth0 Actions for group claims injection, M2M client sync with database-driven groups enrichment for OAuth2 Client Credentials tokens, and complete Docker Compose and Terraform/ECS deployment support. Switch identity providers with a single environment variable while using the same management APIs and UI. Auth0 Setup Guide

• 🔑 Okta Identity Provider Support - Full enterprise SSO integration with Okta as an identity provider. The existing harmonized IAM API now supports Okta alongside Keycloak and Microsoft Entra ID, providing a unified interface to create users, groups, and M2M service accounts regardless of your IdP choice. Features include custom authorization server support for scalable M2M authentication, database-driven groups enrichment for OAuth2 Client Credentials tokens, and complete Docker Compose and Terraform/ECS deployment support. Switch identity providers with a single environment variable while using the same management APIs and UI. Okta Setup Guide

• 🔐 Enterprise Security Posture Documentation - Comprehensive security architecture documentation covering defense-in-depth across all deployment platforms (ECS, EKS, Docker Compose). Details infrastructure security, encryption at rest/in-transit with KMS, secrets management with automated rotation, container hardening following CIS benchmarks, application security with automated scanning (Semgrep, Bandit), supply chain security for MCP servers, and compliance with SOC 2/GDPR standards. Security Posture Guide

• 📊 Direct OTLP Push Export for Metrics - Push metrics directly to any OTLP-compatible observability platform (Datadog, New Relic, Honeycomb, Grafana Cloud) without requiring an intermediate OTEL Collector. Configure via environment variables (OTEL_OTLP_ENDPOINT, OTEL_EXPORTER_OTLP_HEADERS) for instant integration with commercial observability platforms. Supports both Docker Compose and Terraform/ECS deployments with secure credential handling via AWS Secrets Manager. Works alongside existing Prometheus/Grafana setup for hybrid monitoring. Metrics Architecture Guide - Direct OTLP Push

• ⭐ AWS Workshop Studio: Securing AI Agent Ecosystems with MCP Gateway and Registry - Hands-on workshop covering deployment, authentication, governance, and security best practices for production AI agent ecosystems. Learn to deploy the MCP Gateway & Registry on AWS, configure enterprise authentication, implement fine-grained access control, and secure AI agent communications. Start Workshop

• 💻 One-Command macOS Setup - The quickest way to get started and experiment with the solution on your MacBook. Simply ask Claude Code or your favorite AI coding assistant to use the macOS Setup Skill and it will automatically clone the repository, install all dependencies, configure services (MongoDB, Keycloak, registry), register sample servers, and verify the complete stack is running. Perfect for single-developer environments and hands-on exploration. Supports both full setup and complete teardown with a single command. ECS/EKS deployment skill coming very soon.

• AI Registry MCP Server (airegistry-tools) - Enables AI coding assistants (Claude Code, Roo Code, Cursor, etc.) to discover and query MCP servers, agents, and skills directly from the registry. Provides 5 tools: list_services, list_agents, list_skills, intelligent_tool_finder (semantic search), and healthcheck. Auto-registered on registry startup with no manual setup required. See AI Registry Tools documentation for details.

• Governance & Security Enhancements - Enhanced audit logging with searchable filters (username, MCP server) and statistics dashboard showing top users, operations, timeline charts, and per-user activity breakdowns. System uptime and health stats now visible in the header with deployment info, registry statistics, and database status. Comprehensive security hardening via Bandit scanning addressed subprocess security (B603/B607), SQL injection prevention (B608), hardcoded credentials detection (B105), and other vulnerability patterns across the codebase. All security findings documented and resolved with proper justifications for necessary exceptions.

• IAM Settings UI - Visual interface for managing users, groups, and M2M service accounts directly from the web UI. Create and configure access control groups with fine-grained permissions for servers, tools, agents, and UI features. Manage human users with group assignments, and create M2M service accounts for AI agents with OAuth2 client credentials. Features include searchable server/agent/tool selectors, JSON import/export for scope configurations, and support for both MCP servers and virtual servers in access rules. Works with both Keycloak and Microsoft Entra ID identity providers. IAM Settings Guide

• System Configuration Viewer - View and export all registry configuration parameters through the Settings UI. Admin-only panel displays 11 configuration groups (Deployment, Storage, Auth, Embeddings, Health, WebSocket, Security Scanning, Audit, Federation, Discovery) with sensitive value masking. Export configuration in ENV, JSON, TFVARS, or YAML formats for deployment automation. API endpoints provide programmatic access at /api/config/full and /api/config/export. Configuration Guide

• Virtual MCP Server Support - Aggregate tools, resources, and prompts from multiple backend MCP servers into a single unified endpoint. Clients connect to one virtual server that presents a curated, access-controlled view of capabilities from any combination of registered backends. Features include tool aliasing (resolve naming conflicts), version pinning (lock to specific backend versions), per-tool scope-based access control, session multiplexing (one client session maps to N backend sessions transparently), and 60-second cached aggregation for tools/list, resources/list, and prompts/list. Supports all MCP JSON-RPC methods including initialize, ping, tools/call, resources/read, and prompts/get. Design Document | Operations Guide

• Registry-Only Deployment Mode - Run the registry as a standalone catalog/discovery service without nginx gateway integration. In registry-only mode, nginx configuration is not updated when servers are registered, and MCP proxy requests return 503 with instructions to use direct connection. The frontend adapts to show proxy_pass_url instead of gateway URLs. Combined with REGISTRY_MODE settings (full, skills-only, mcp-servers-only, agents-only), you can configure the registry for specific use cases. For example, set REGISTRY_MODE=skills-only to run a dedicated Skills Registry that only manages Agent Skills (SKILL.md files) without MCP servers or A2A agents - ideal for teams that want a lightweight skill library. The UI automatically adapts to show only relevant features, and API endpoints for disabled features return 503. Invalid combinations like with-gateway + skills-only are auto-corrected with warnings. Registry Deployment Modes Guide

• Agent Skills Registry - Register, discover, and manage reusable instruction sets (SKILL.md files) that enhance AI coding assistants with specialized workflows. Skills are hosted on GitHub, GitLab, or Bitbucket and registered in the MCP Gateway Registry for discovery and access control. Features include YAML frontmatter parsing for metadata extraction, health monitoring with URL accessibility checks, visibility controls (public/private/group), star ratings, semantic search integration, tool dependency validation, and a rich UI with SKILL.md content modals. Security includes automatic security scanning during registration using Cisco AI Defense Skill Scanner with YARA pattern matching, LLM analysis, and static code inspection. SSRF protection with redirect validation ensures safe URL handling. Agent Skills Guide | Architecture | Security Scanning

• 📋 Compliance Audit Logging - Comprehensive audit logging for security monitoring and compliance. Captures all Registry API and MCP Gateway access events with user identity, operation details, and timing. Features include automatic credential masking (tokens, cookies, passwords are never logged), TTL-based log retention (default 7 days, configurable), admin-only audit viewer UI with filtering and export (JSONL/CSV), and non-blocking async design. Supports SOC 2 and GDPR requirements with who/what/when/where/outcome tracking. Audit Logging Guide

• 🌐 Peer-to-Peer Registry Federation - Connect multiple MCP Gateway Registry instances for bidirectional server and agent synchronization. Central IT teams can aggregate visibility across Line of Business registries, or LOBs can inherit shared tools from a central hub. Features include configurable sync modes (all, whitelist, tag filter), scheduled and on-demand sync, static token authentication for IdP-agnostic deployments, Fernet-encrypted credential storage, generation-based orphan detection, and path namespacing to prevent collisions. Synced items are read-only and display their source registry. A VS Code-style Settings UI provides peer management, sync triggering, and status monitoring. Architecture Design | Operational Guide

• 🔑 Static Token Auth for Registry API - Access Registry API endpoints (/api/*, /v0.1/*) using a static API key instead of IdP-based JWT validation. Designed for trusted network environments, CI/CD pipelines, and CLI tooling where configuring a full identity provider may not be practical. MCP Gateway endpoints continue to require full IdP authentication. Includes startup validation that disables the feature if no token is configured. Registry API Authentication Guide

• 🔀 MCP Server Version Routing - Run multiple versions of the same MCP server simultaneously behind a single gateway endpoint. Register new versions as inactive, test them with the X-MCP-Server-Version header, then promote to active with a single API call or UI click. Features include instant rollback, version pinning for clients, deprecation lifecycle with sunset dates, automatic nginx map-based O(1) routing, cascade deletion of all versions, and post-swap health checks. The dashboard displays both the admin-controlled routing version and the MCP server-reported software version independently. Only the active version appears in search results and health checks. Design Document | Operations Guide

• 👥 Multi-Provider IAM with Harmonized API - Full Identity and Access Management support for Keycloak, Microsoft Entra ID, Okta, and Auth0. The registry API provides a unified experience for user and group management regardless of which IdP you use. Human users can log in via the UI and generate self-signed JWT tokens (with the same permissions as their session) for CLI tools and AI coding assistants. Service accounts (M2M) enable AI agent identity with OAuth2 Client Credentials flow. Fine-grained access control through scopes defines exactly which MCP servers, methods, tools, and agents each user can access. Authentication Design | IdP Provider Architecture | Scopes Management | Entra ID Setup | Okta Setup | Auth0 Setup

• 🏷️ Custom Metadata for Servers & Agents - Add rich custom metadata to MCP servers and agents for organization, compliance, and integration tracking. Metadata is fully searchable via semantic search, enabling queries like "team:data-platform", "PCI-DSS compliant", or "owner:alice@example.com". Use cases include team ownership, compliance tracking (PCI-DSS, HIPAA), cost center allocation, deployment regions, JIRA tickets, and custom tags. Backward compatible with existing registrations. Metadata Usage Guide

• 🔎 Enhanced Hybrid Search - Improved semantic search combining vector similarity with tokenized keyword matching for servers, tools, and agents. Explicit name references now boost relevance scores, ensuring exact matches appear first. Hybrid Search Architecture

• 🛡️ Security Scan Results in UI - Security scan results are now displayed directly on Server and Agent cards with color-coded shield icons (gray/green/red). Click the shield icon to view detailed scan results and trigger rescans from the UI. Security Scanner Documentation

• 🧪 Comprehensive Test Suite & Updated LLM Documentation - Full pytest test suite with 701+ passing tests (unit, integration, E2E) running automatically on all PRs via GitHub Actions. 35% minimum coverage (targeting 80%), ~30 second execution with 8 parallel workers. Updated llms.txt provides comprehensive documentation for LLM coding assistants covering storage backend migration (file → DocumentDB/MongoDB), repository patterns, AWS ECS deployment, Microsoft Entra ID integration, dual security scanning, federation architecture, rating system, testing standards, and critical code organization antipatterns. Testing Guide | docs/llms.txt

• 📊 DocumentDB & MongoDB CE Storage Backend - Distributed storage with MongoDB-compatible backends. DocumentDB provides native HNSW vector search for sub-100ms semantic queries in production deployments, while MongoDB Community Edition 8.2 enables full-featured local development with replica sets. Both backends use the same repository abstraction layer with automatic collection management, optimized indexes, and application-level vector search for MongoDB CE. Switch between MongoDB CE (local testing) and DocumentDB (production) with a single environment variable. Note: File-based storage is deprecated and will be removed in a future release. MongoDB CE is recommended for local development. Configuration Guide | Storage Architecture

• 🔒 A2A Agent Security Scanning - Integrated security scanning for A2A agents using Cisco AI Defense A2A Scanner. Automatic security scans during agent registration with YARA pattern matching, A2A specification validation, and heuristic threat detection. Features include automatic tagging of unsafe agents, configurable blocking policies, and detailed scan reports with API endpoints for viewing results and triggering rescans.

• 🔧 Registry Management API - New programmatic API for managing servers, groups, and users. Python client (api/registry_client.py) with type-safe interfaces, RESTful HTTP endpoints (/api/management/*), and comprehensive error handling. Replaces shell scripts with modern API approach while maintaining backward compatibility. API Documentation | Service Management Guide

• ⭐ Server & Agent Rating System - Rate and review agents with an interactive 5-star rating widget. Users can submit ratings via the UI or CLI, view aggregate ratings with individual rating details, and update their existing ratings. Features include a rotating buffer (max 100 ratings per agent), one rating per user, float average calculations, and full OpenAPI documentation. Enables community-driven agent quality assessment and discovery.

• 🧠 Flexible Embeddings Support - Choose from three embedding provider options for semantic search: local sentence-transformers, OpenAI, or any LiteLLM-supported provider including Amazon Bedrock Titan, Cohere, and 100+ other models. Switch providers with simple configuration changes. Embeddings Guide

• ☁️ AWS ECS Deployment - Deployment configuration on Amazon ECS Fargate with multi-AZ architecture, Application Load Balancer with HTTPS, auto-scaling, CloudWatch monitoring, and NAT Gateway redundancy. Complete Terraform configuration for deploying the entire stack. ECS Deployment Guide

• 📦 Flexible Deployment Modes - Three deployment options to match your requirements: (1) CloudFront Only for quick setup without custom domains, (2) Custom Domain with Route53/ACM for branded URLs, or (3) CloudFront + Custom Domain for production with CDN benefits. Deployment Modes Guide

• 🔗 Federated Registry - MCP Gateway registry now supports federation of servers and agents from other registries. Federation Guide

• 🔗 Agent-to-Agent (A2A) Protocol Support - Agents can now register, discover, and communicate with other agents through a secure, centralized registry. Enable autonomous agent ecosystems with Keycloak-based access control and fine-grained permissions. A2A Guide

• 🏢 Microsoft Entra ID Integration - Enterprise SSO with Microsoft Entra ID (Azure AD) authentication. Group-based access control, conditional access policies, and seamless integration with existing Microsoft 365 environments. Entra ID Setup Guide

• 🤖 Agentic CLI for MCP Registry - Talk to the Registry in natural language using a Claude Code-like interface. Discover tools, ask questions, and execute MCP commands conversationally. Learn more

• 🔒 MCP Server Security Scanning - Integrated vulnerability scanning with Cisco AI Defense MCP Scanner. Automatic security scans during server registration, periodic registry-wide scans with detailed markdown reports, and automatic disabling of servers with security issues.

• 📥 Import Servers from Anthropic MCP Registry - Import curated MCP servers from Anthropic's registry with a single command. Import Guide

• 🔌 Anthropic MCP Registry REST API Compatibility - Full compatibility with Anthropic's MCP Registry REST API specification. API Documentation

• 🔎 Unified Semantic Search for Servers, Tools & Agents - Natural-language search across every MCP server, its tools, and registered A2A agents using POST /api/search/semantic. Works from the dashboard UI (session cookie auth) or programmatically with JWT Bearer tokens, returning relevance-scored matches per entity type in a single response.

• 🚀 Pre-built Images - Deploy instantly with pre-built Docker images. Get Started | macOS Guide

• 🔐 Keycloak Integration - Enterprise authentication with AI agent audit trails and group-based authorization. Learn more

• ⚡ Amazon Bedrock AgentCore Integration - AgentCore Gateway support with dual authentication. Integration Guide

---

A2A Agents - Example Implementations

The registry includes two example A2A agents that demonstrate how both human developers and autonomous AI agents can discover, register, and use agents through the unified Agent Registry. Agents can programmatically discover other agents via semantic search and use them through the A2A protocol, enabling dynamic agent composition and autonomous agent orchestration.

Example Agents

Agent	Path	Skills
Travel Assistant Agent	/travel-assistant-agent	Flight search, pricing checks, recommendations, trip planning
Flight Booking Agent	/flight-booking-agent	Availability checks, flight reservations, payments, reservation management

Agent Discovery

View in Registry UI: Open the registry and navigate to the A2A Agents tab to browse registered agents with their full metadata, capabilities, and skills.

Search via CLI: Developers can search for agents by natural language description:

# Search for agents that can help book a trip
cli/agent_mgmt.sh search "need an agent to book a trip"

Example Output:

Found 4 agent(s) matching 'need an agent to book a trip':
--------------------------------------------------------------------------------------------------------------
Agent Name                               | Path                      | Score
--------------------------------------------------------------------------------------------------------------
Travel Assistant Agent                   | /travel-assistant-agent   |  0.8610
Flight Booking Agent                     | /flight-booking-agent     |  1.2134
--------------------------------------------------------------------------------------------------------------

Agent-to-Agent Discovery API

The registry provides a semantic search API that agents can use as a tool to discover other A2A agents at runtime. This API enables dynamic agent composition where agents find collaborators based on capabilities rather than hardcoded references.

Discovery API Endpoint:

POST /api/agents/discover/semantic?query=<natural-language-query>&max_results=5
Authorization: Bearer <jwt-token>

Response includes:

• Agent name, description, and endpoint URL
• Agent card metadata with skills and capabilities
• Relevance score for ranking matches
• Trust level and visibility settings

How agents use it:

1. An agent calls the registry's semantic search API with a natural language query (e.g., "agent that can book flights")
2. The registry returns matching agents with their endpoint URLs and full agent card metadata
3. The agent uses the agent card to understand capabilities and invokes the discovered agent via A2A protocol

Example - Travel Assistant discovering and invoking Flight Booking Agent:

User: "I need to book a flight from NYC to LA"

Travel Assistant:
  1. Calls registry API: POST /api/agents/discover/semantic?query="book flights"
  2. Registry returns Flight Booking Agent with endpoint URL and agent card
  3. Uses agent card to understand capabilities, then sends A2A message to Flight Booking Agent
  4. Returns booking confirmation to user

This pattern enables agents to dynamically extend their capabilities by discovering specialized agents for tasks they cannot handle directly.

Agent Cards: View the agent card metadata at agents/a2a/test/ to see the complete agent definitions including skills, protocols, and capabilities.

For complete agent deployment and testing documentation, see agents/a2a/README.md.

---

Core Use Cases

AI Agent & Coding Assistant Governance

Provide both autonomous AI agents and human developers with secure access to approved tools through AI coding assistants (VS Code, Cursor, Claude Code) while maintaining IT oversight and compliance.

Enterprise Security & Compliance

Centralized authentication, fine-grained permissions, and comprehensive audit trails for SOX/GDPR compliance pathways across both human and AI agent access patterns.

Dynamic Tool Discovery

AI agents can autonomously discover and execute specialized tools beyond their initial capabilities using intelligent semantic search, while developers get guided tool discovery through their coding assistants.

Unified Access Gateway

Single gateway supporting both autonomous AI agents (machine-to-machine) and AI coding assistants (human-guided) with consistent authentication and tool access patterns.

---

Architecture

The MCP Gateway & Registry provides a unified platform for both autonomous AI agents and AI coding assistants to access enterprise-curated tools through a centralized gateway with comprehensive authentication and governance.

flowchart TB
    subgraph Human_Users["Human Users"]
        User1["Human User 1"]
        User2["Human User 2"]
        UserN["Human User N"]
    end

    subgraph AI_Agents["AI Agents"]
        Agent1["AI Agent 1"]
        Agent2["AI Agent 2"]
        Agent3["AI Agent 3"]
        AgentN["AI Agent N"]
    end

    subgraph EC2_Gateway["<b>MCP Gateway & Registry</b> (Amazon EC2 Instance)"]
        subgraph NGINX["NGINX Reverse Proxy"]
            RP["Reverse Proxy Router"]
        end
        
        subgraph AuthRegistry["Authentication & Registry Services"]
            AuthServer["Auth Server<br/>(Dual Auth)"]
            Registry["Registry<br/>Web UI"]
            RegistryMCP["Registry<br/>MCP Server"]
        end
        
        subgraph LocalMCPServers["Local MCP Servers"]
            MCP_Local1["MCP Server 1"]
            MCP_Local2["MCP Server 2"]
        end
    end
    
    %% Identity Provider
    IdP[Identity Provider<br/>Keycloak/Cognito]
    
    subgraph EKS_Cluster["Amazon EKS/EC2 Cluster"]
        MCP_EKS1["MCP Server 3"]
        MCP_EKS2["MCP Server 4"]
    end
    
    subgraph APIGW_Lambda["Amazon API Gateway + AWS Lambda"]
        API_GW["Amazon API Gateway"]
        Lambda1["AWS Lambda Function 1"]
        Lambda2["AWS Lambda Function 2"]
    end
    
    subgraph External_Systems["External Data Sources & APIs"]
        DB1[(Database 1)]
        DB2[(Database 2)]
        API1["External API 1"]
        API2["External API 2"]
        API3["External API 3"]
    end
    
    %% Connections from Human Users
    User1 -->|Web Browser<br>Authentication| IdP
    User2 -->|Web Browser<br>Authentication| IdP
    UserN -->|Web Browser<br>Authentication| IdP
    User1 -->|Web Browser<br>HTTPS| Registry
    User2 -->|Web Browser<br>HTTPS| Registry
    UserN -->|Web Browser<br>HTTPS| Registry
    
    %% Connections from Agents to Gateway
    Agent1 -->|MCP Protocol<br>SSE with Auth| RP
    Agent2 -->|MCP Protocol<br>SSE with Auth| RP
    Agent3 -->|MCP Protocol<br>Streamable HTTP with Auth| RP
    AgentN -->|MCP Protocol<br>Streamable HTTP with Auth| RP
    
    %% Auth flow connections
    RP -->|Auth validation| AuthServer
    AuthServer -.->|Validate credentials| IdP
    Registry -.->|User authentication| IdP
    RP -->|Tool discovery| RegistryMCP
    RP -->|Web UI access| Registry
    
    %% Connections from Gateway to MCP Servers
    RP -->|SSE| MCP_Local1
    RP -->|SSE| MCP_Local2
    RP -->|SSE| MCP_EKS1
    RP -->|SSE| MCP_EKS2
    RP -->|Streamable HTTP| API_GW
    
    %% Connections within API GW + Lambda
    API_GW --> Lambda1
    API_GW --> Lambda2
    
    %% Connections to External Systems
    MCP_Local1 -->|Tool Connection| DB1
    MCP_Local2 -->|Tool Connection| DB2
    MCP_EKS1 -->|Tool Connection| API1
    MCP_EKS2 -->|Tool Connection| API2
    Lambda1 -->|Tool Connection| API3

    %% Style definitions
    classDef user fill:#fff9c4,stroke:#f57f17,stroke-width:2px
    classDef agent fill:#e1f5fe,stroke:#29b6f6,stroke-width:2px
    classDef gateway fill:#e8f5e9,stroke:#66bb6a,stroke-width:2px
    classDef nginx fill:#f3e5f5,stroke:#ab47bc,stroke-width:2px
    classDef mcpServer fill:#fff3e0,stroke:#ffa726,stroke-width:2px
    classDef eks fill:#ede7f6,stroke:#7e57c2,stroke-width:2px
    classDef apiGw fill:#fce4ec,stroke:#ec407a,stroke-width:2px
    classDef lambda fill:#ffebee,stroke:#ef5350,stroke-width:2px
    classDef dataSource fill:#e3f2fd,stroke:#2196f3,stroke-width:2px
    
    %% Apply styles
    class User1,User2,UserN user
    class Agent1,Agent2,Agent3,AgentN agent
    class EC2_Gateway,NGINX gateway
    class RP nginx
    class AuthServer,Registry,RegistryMCP gateway
    class IdP apiGw
    class MCP_Local1,MCP_Local2 mcpServer
    class EKS_Cluster,MCP_EKS1,MCP_EKS2 eks
    class API_GW apiGw
    class Lambda1,Lambda2 lambda
    class DB1,DB2,API1,API2,API3 dataSource

Key Architectural Benefits:

• Unified Gateway: Single point of access for both AI agents and human developers through coding assistants
• Dual Authentication: Supports both human user authentication and machine-to-machine agent authentication
• Scalable Infrastructure: Nginx reverse proxy with horizontal scaling capabilities
• Multiple Transports: SSE and Streamable HTTP support for different client requirements

---

Key Advantages

Security Features

• OAuth 2.0/3.0 compliance with IdP integration
• Fine-grained access control at tool and method level
• Zero-trust network architecture
• Complete audit trails and comprehensive analytics for compliance

AI Agent & Developer Experience

• Single configuration works across autonomous AI agents and AI coding assistants (VS Code, Cursor, Claude Code, Cline)
• Dynamic tool discovery with natural language queries for both agents and humans
• Instant onboarding for new team members and AI agent deployments
• Unified governance for both AI agents and human developers

Deployment Features

• Container-native (Docker/Kubernetes)
• Real-time health monitoring and alerting
• Dual authentication supporting both human and machine authentication

---

Quick Start

There are 4 options for setting up the MCP Gateway & Registry:

• Option A: AI-Assisted macOS Setup — The absolute fastest way to get started on macOS. Ask your AI coding assistant to use the macOS Setup Skill for fully automated one-command setup. Perfect for experimentation.
• Option B: Pre-built Images — Fast setup using pre-built Docker or Podman containers. Recommended for most users.
• Option C: Podman (Rootless) — Detailed Podman-specific instructions for macOS and rootless Linux environments.
• Option D: Build from Source — Full source build for customization or development.

Option A: AI-Assisted macOS Setup (Fastest)

The easiest way to get started on macOS. Simply ask Claude Code or your AI coding assistant:

"Use the macOS setup skill to install and configure the MCP Gateway & Registry"

The macOS Setup Skill will automatically:

• ✅ Clone the repository and install all dependencies (Homebrew, Python, UV, Docker, Node.js)
• ✅ Configure and start MongoDB with replica set
• ✅ Set up and initialize Keycloak with admin user
• ✅ Start the registry and auth server
• ✅ Register the Cloudflare MCP docs server
• ✅ Verify the complete stack is operational

Perfect for: Single-developer experimentation, quick demos, hands-on exploration

What you need: macOS with an AI coding assistant (Claude Code, Cursor, etc.)

Clean up: When done, ask your AI assistant to "teardown the MCP Gateway setup" for complete removal.

Note: ECS/EKS deployment skill coming very soon for production deployments.

---

Option B: Pre-built Images (Instant Setup)

Get running with pre-built Docker containers in minutes. This is the recommended approach for most users.

# Clone and configure
git clone https://github.com/agentic-community/mcp-gateway-registry.git
cd mcp-gateway-registry
cp .env.example .env

# Edit .env with your passwords (KEYCLOAK_ADMIN_PASSWORD, etc.)
nano .env

# Deploy with pre-built images
export DOCKERHUB_ORG=mcpgateway
./build_and_run.sh --prebuilt

# Access the Registry UI
open http://localhost:7860  # macOS
# xdg-open http://localhost:7860  # Linux

Complete Quick Start Guide - Full step-by-step instructions including:

• Prerequisites installation (Docker, Python, UV)
• Environment configuration
• MongoDB and Keycloak initialization
• User and service account setup
• Server and agent registration
• Testing the gateway functionality

Benefits: No build time | No Node.js required | No frontend compilation | Consistent tested images

---

Option C: Podman (Rootless Container Deployment)

Perfect for macOS and rootless Linux environments

Podman provides rootless container execution without requiring privileged ports, making it ideal for:

• macOS users with Podman Desktop
• Linux users preferring rootless containers
• Development environments where Docker daemon isn't available

Quick Podman Setup (macOS non-Apple Silicon):

# Install Podman Desktop
brew install podman-desktop
# OR download from: https://podman-desktop.io/

Inside Podman Desktop, go to Preferences > Podman Machine and create a new machine with at least 4 CPUs and 8GB RAM. Alternatively, see more detailed Podman installation guide for instructions on setting this up on CLI.

# Initialize Podman machine
podman machine init
podman machine start

# Verify installation
podman --version
podman compose version

# Configure environment
cp .env.example .env
# Edit .env with your credentials

Deploy with Podman see full Podman setup instructions (downloading, installing, and initializing a first Podman container, as well as troubleshooting) in our Installation Guide.

Build with Podman:

# Auto-detect (will use Podman if Docker not available)
./build_and_run.sh --prebuilt

# Explicit Podman mode (only non-Apple Silicon)
./build_and_run.sh --prebuilt --podman

# Access registry at non-privileged ports
# On macOS:
open http://localhost:8080
# On Linux: xdg-open http://localhost:8080

Note: Apple Silicon (M1/M2/M3)? Don't use --prebuilt with Podman on ARM64. This will cause a "proxy already running" error. See Podman on Apple Silicon Guide.

# To run on Apple Silicon Macs:
./build_and_run.sh --podman

Key Differences vs. Docker:

• No root/sudo required
• Works on macOS without privileged port access
• HTTP port: 8080 (instead of 80)
• HTTPS port: 8443 (instead of 443)
• All other service ports unchanged

For detailed Podman setup instructions, see Installation Guide and macOS Setup Guide.

Option D: Build from Source

New to MCP Gateway? Start with our Complete Setup Guide for detailed step-by-step instructions from scratch on AWS EC2.

Running on macOS? See our macOS Setup Guide for platform-specific instructions and optimizations.

Testing & Integration Options

Test Suite: The project includes comprehensive automated testing with pytest:

# Run all tests
make test

# Run only unit tests (fast)
make test-unit

# Run with coverage report
make test-coverage

# Run specific test categories
uv run pytest -m unit           # Unit tests only
uv run pytest -m integration    # Integration tests
uv run pytest -m "not slow"     # Skip slow tests

Test Structure:

• Unit Tests (tests/unit/) - Fast, isolated component tests
• Integration Tests (tests/integration/) - Component interaction tests
• E2E Tests (tests/integration/test_e2e_workflows.py) - Complete workflow tests

Python Agent:

• agents/agent.py - Full-featured Python agent with advanced AI capabilities

Testing Documentation:

• Testing Guide - Comprehensive testing documentation
• Writing Tests - How to write effective tests
• Test Maintenance - Maintaining test suite health

Pre-commit Hooks:

# Install pre-commit hooks
pip install pre-commit
pre-commit install

# Run hooks manually
pre-commit run --all-files

Next Steps: Complete Installation Guide | Authentication Setup | AI Assistant Integration

---

Enterprise Features

AI Agents & Coding Assistants Integration

Transform how both autonomous AI agents and development teams access enterprise tools with centralized governance:

Enterprise-curated MCP servers accessible through unified gateway

AI assistants executing approved enterprise tools with governance

Observability Comprehensive real-time metrics and monitoring through Grafana dashboards with dual-path storage: SQLite for detailed historical analysis and OpenTelemetry (OTEL) export for integration with Prometheus, CloudWatch, Datadog, and other monitoring platforms. Track authentication events, tool executions, discovery queries, and system performance metrics. Learn more Real-time metrics and observability dashboard tracking server health, tool usage, and authentication events

Anthropic MCP Registry Integration

Seamlessly integrate with Anthropic's official MCP Registry to import and access curated MCP servers through your gateway:

• Import Servers: Select and import desired servers from Anthropic's registry with a single command
• Unified Access: Access imported servers through your gateway with centralized authentication and governance
• API Compatibility: Full support for Anthropic's Registry REST API specification - point your Anthropic API clients to this registry to discover available servers

Import and access curated MCP servers from Anthropic's official registry

Import Guide | Registry API Documentation

Federation - External Registry Integration

Unified Multi-Registry Access:

• Anthropic MCP Registry - Import curated MCP servers with purple ANTHROPIC visual tags
• Workday ASOR - Import AI agents from Agent System of Record with orange ASOR visual tags
• Automatic Sync - Scheduled synchronization with external registries
• Visual Identification - Clear visual tags distinguish federation sources in the UI
• Centralized Management - Single control plane for all federated servers and agents

Quick Setup:

# Configure federation sources
echo 'ASOR_ACCESS_TOKEN=your_token' >> .env

# Update federation.json with your sources
# Restart services
./build_and_run.sh

📖 Complete Federation Guide - Environment setup, authentication, configuration, and troubleshooting

Security Scanning

Integrated Vulnerability Detection:

• Automated Security Scanning - Integrated vulnerability scanning for MCP servers using Cisco AI Defence MCP Scanner, with automatic scans during registration and support for periodic registry-wide scans
• Detailed Security Reports - Comprehensive markdown reports with vulnerability details, severity assessments, and remediation recommendations
• Automatic Protection - Servers with security issues are automatically disabled with security-pending status to protect your infrastructure
• Compliance Ready - Security audit trails and vulnerability tracking for enterprise compliance requirements

Authentication & Authorization

Multiple Identity Modes:

• Machine-to-Machine (M2M) - For autonomous AI agents and automated systems
• Three-Legged OAuth (3LO) - For external service integration (Atlassian, Google, GitHub)
• Session-Based - For human developers using AI coding assistants and web interface

Supported Identity Providers: Keycloak, Microsoft Entra ID, Okta, Auth0, Amazon Cognito, and any OAuth 2.0 compatible provider. Learn more

Fine-Grained Permissions: Tool-level, method-level, team-based, and temporary access controls. Learn more

Deployment Options

Cloud Platforms: Amazon EC2, Amazon EKS

---

Telemetry

The registry collects anonymous, non-sensitive usage telemetry to help us understand adoption patterns and improve the product. Both tiers are opt-out and on by default.

What is sent (Tier 1 -- startup ping): Registry version, Python version, OS, CPU architecture, cloud provider, storage backend, auth provider, deployment mode, embeddings provider (sentence-transformers or litellm), and embeddings backend kind (a derived coarse category: sentence-transformers, bedrock, openai, azure-openai, voyage, cohere, other, or unknown). No IP addresses, hostnames, file paths, user data, or any PII.

Also sent by default (Tier 2 -- daily heartbeat): Aggregate counts (number of servers, agents, skills, peers), search backend, embeddings provider, embeddings backend kind, and uptime. Same privacy guarantees as Tier 1. Disable heartbeat only: MCP_TELEMETRY_OPT_OUT=1.

What is never sent: The raw EMBEDDINGS_MODEL_NAME, EMBEDDINGS_MODEL_DIMENSIONS, credentials (API keys, AWS secrets), endpoints (EMBEDDINGS_API_BASE), and regions (EMBEDDINGS_AWS_REGION). Only the derived backend-kind category is reported. See docs/TELEMETRY.md for the complete schema and privacy guarantees.

Behavior change (post v1.0.18): The daily heartbeat was previously opt-in (MCP_TELEMETRY_OPT_IN=1). It is now opt-out and sent by default. Since the heartbeat contains only aggregate counts (no PII), this aligns it with the startup ping behavior.

To opt out completely:

export MCP_TELEMETRY_DISABLED=1   # Disables both startup ping and heartbeat

To disable heartbeat only (startup ping still sent):

export MCP_TELEMETRY_OPT_OUT=1

All requests are HMAC-signed, rate-limited, and schema-validated. Telemetry is fail-silent and never impacts registry operation. Full details in the Telemetry Documentation.

---

Deployments

AWS Elastic Container Service (ECS)

Deployment configuration on Amazon ECS Fargate with comprehensive enterprise features:

• Multi-AZ Architecture - Redundancy across multiple availability zones
• Application Load Balancer - HTTPS/SSL termination with automatic certificate management via ACM
• Auto-scaling - Dynamic scaling based on CPU and memory utilization
• CloudWatch Integration - Comprehensive monitoring, logging, and alerting
• NAT Gateway HA - Redundant NAT gateway configuration for secure outbound connectivity
• Keycloak Integration - Enterprise authentication with RDS Aurora PostgreSQL backend
• EFS Shared Storage - Persistent storage for models, logs, and configuration
• Service Discovery - AWS Cloud Map for service-to-service communication

Complete ECS Deployment Guide - Step-by-step instructions for deploying the entire stack with Terraform.

Amazon EKS (Kubernetes)

Coming Soon - Kubernetes deployment on Amazon EKS with Helm charts for container orchestration at scale.

---

Documentation

Getting Started	Enterprise Setup	Developer & Operations
Complete Setup Guide NEW! Step-by-step from scratch on AWS EC2	Authentication Guide OAuth and identity provider integration	AI Coding Assistants Setup VS Code, Cursor, Claude Code integration
Installation Guide Complete setup instructions for EC2 and EKS	AWS ECS Deployment Deployment guide for AWS ECS Fargate	API Reference Programmatic registry management
Keycloak Integration Enterprise identity with agent audit trails	Token Refresh Service Automated token refresh and lifecycle management	MCP Registry CLI Command-line client for registry management
Configuration Reference Environment variables and settings	Amazon Cognito Setup Step-by-step IdP configuration	Observability Guide NEW! Metrics, monitoring, and OpenTelemetry setup
Auth0 Integration Auth0 SSO with M2M support	Okta Setup Okta IdP configuration	Entra ID Setup Microsoft Entra ID integration
	Anthropic Registry Import NEW! Import servers from Anthropic MCP Registry	Federation Guide External registry integration (Anthropic, ASOR)
		P2P Federation Guide NEW! Peer-to-peer registry federation
	Service Management Server lifecycle and operations	Anthropic Registry API NEW! REST API compatibility
		Fine-Grained Access Control Permission management and security
		Dynamic Tool Discovery Autonomous agent capabilities
		Deployment Guide Complete setup for deployment environments
		Troubleshooting Guide Common issues and solutions

---

Community

Get Involved

Join the Discussion

• GitHub Discussions - Feature requests and general discussion
• GitHub Issues - Bug reports and feature requests

Contributing

• Contributing Guide - How to contribute code and documentation
• Code of Conduct - Community guidelines
• Security Policy - Responsible disclosure process

Star History

Roadmap

Our development roadmap is organized into release milestones with clear deliverables and progress tracking:

Milestone	Progress	Status	Key Issues
v1.0.20	100% (11/11)	Complete	#871 - Unified Auth, #851 - M2M Registration, #824 - Python 3.14, #809 - Registration Gate, #779 - Multi API Keys, #742 - Webhooks and 5 more
v1.0.21	100% (5/5)	Complete	#906 - Admin Data Export, #897 - Per-skill Auth Credentials UI, #891 - CSRF Toggle Fix, #886 - Centralized Log Rotation, #856 - ARM64 Images
v1.0.22	0% (0/5)	Planned	#867 - Prometheus Metrics Endpoint, #847 - A2A Reverse Proxy Gateway, #844 - Dependency Management, #744 - AI Chat Assistant, #500 - Logout Routing Fix
Parking Lot	Backlog	Backlog	23 open issues awaiting prioritization

Status Legend: Complete, Planned, Backlog

---

Major Features

The following major features span multiple milestones and represent significant architectural improvements:

• #867 - Prometheus Metrics Endpoint PLANNED (v1.0.22) Expose a /metrics endpoint on the registry for in-process Prometheus counters.

• #847 - A2A Reverse Proxy Gateway PLANNED (v1.0.22) Add reverse proxy gateway support for A2A agents.

• #744 - AI Chat Assistant PLANNED (v1.0.22) Embedded AI chat assistant for registry operations, discovery, and agent design.

• #665 - Agent-to-Agent Knowledge Sharing BACKLOG Enable agents to share and discover knowledge through the AI Registry, forming a collaborative knowledge network.

• #666 - Context Hub MVP BACKLOG Implement Context Hub with card creation, search, and auto-discovery for agent knowledge management.

• #614 - MCP OAuth 2.1 Authorization Spec BACKLOG Implement RFC 9728 Protected Resource Metadata with native IDE support for MCP OAuth 2.1 authorization.

• #556 - AI Gateway & Registry Rebrand BACKLOG Rename "MCP Gateway Registry" to "AI Gateway & Registry" to reflect expanded support for agents and tools beyond MCP.

• #605 - AgentCore Auto-Registration COMPLETED (April 2026) Automated discovery and registration of Bedrock AgentCore gateways with credential management integration. Full cli/agentcore/ module with boto3 discovery, registration, token refresh, and security scheme support.

• #641 - Okta Identity Provider COMPLETED Added Okta as an identity provider option alongside Keycloak, Entra ID, Auth0, GitHub, and Google OAuth2.

• #557-559 - Observability & Telemetry Suite COMPLETED Comprehensive telemetry infrastructure with server-side collector, client-side instrumentation, and end-to-end enhancements. Telemetry docs.

• #129 - Virtual MCP Server Support COMPLETED Dynamic tool aggregation and intelligent routing using Lua scripting. Enables logical grouping of tools from multiple backend servers into a single virtual endpoint.

• #232 - A2A Curated Registry Discovery COMPLETED Enable agent-to-agent discovery and tool invocation through curated registry patterns.

• #260 - Federation Between MCP Registry Instances COMPLETED Federated registry with bi-directional sync, peer management, chain prevention, orphan detection, and security scan propagation across registries.

• #297 - Unified UI Registration Flow COMPLETED Streamlined registration experience for both MCP servers and A2A agents through a unified interface.

• #295 - Multi-Level Tool Usage Rate Limiting BACKLOG Comprehensive rate limiting architecture with detailed implementation guide for tool usage control.

---

Recently Completed (April 2026)

• #906 - Admin Data Export - Admin-only Data Export page for downloading registry collections as JSON.
• #897 - Per-skill Auth Credentials UI - Frontend UI for managing per-skill authentication credentials.
• #886 - Centralized Log Rotation - Centralized log rotation, auth-server file logging, and log retrieval via MongoDB storage.
• #871 - Unified JWT and Static Token Auth - JWT/session auth coexists with static token auth, supporting four credential types concurrently.
• #856 - ARM64 Docker Images - Multi-architecture Docker images with ARM64 support.
• #851 - Direct M2M Client Registration - Direct machine-to-machine client registration API that bypasses IdP sync.
• #824 - Python 3.14 Runtime Upgrade - Upgraded Python runtime from 3.12 to 3.14 to resolve CVE-2025-13836.
• #809 - Registration Gate Admission Control - Admission control webhook for agent, server, and skill registration.
• #779 - Multiple Static API Keys - Multiple static API keys with per-key group and scope assignments.
• #742 - Webhook Notifications - Configurable webhook notification on server, agent, and skill registration events.

For the complete list of all issues, feature requests, and detailed release history, visit:

• All GitHub Issues
• All GitHub Milestones
• Release Notes

---

License

This project is licensed under the Apache-2.0 License - see the LICENSE file for details.

---