agentregistry

A centralized registry to securely curate, discover, deploy, and manage agentic infrastructure from MCP servers, agents to skills.

What is Agent Registry?

Agent Registry brings governance and control to AI artifacts and infrastructure, empowering developers to quickly build and deploy AI applications with confidence. It provides a secure, centralized registry where teams can publish, discover, and share AI artifacts, including MCP servers, agents, and skills, and deploy them seamlessly to any environment.

Agent Registry provides:

• 📦 Centralized Registry: Package, discover and curate AI artifacts from a central source
• 🔒 Control and Governance: Selectively and control custom collection of artifacts
• 📊 Data Enrichment: Automatically validate and score ingested data for insights
• 🌐 Unify AI Infrastructure: Deploy and access artifacts anywhere

See it in action

Learn how to create an Anthropic Skill, publish it to agentregistry, and use it in Claude Code

Agent Registry Architecture

For Operators: Enrich, package, curate and deploy with control

For Developers: Build, push, pull and run applications with confidence

Development setup

See DEVELOPMENT.md for detailed architecture information.

🚀 Quick Start

Prerequisites

• Docker Desktop with Docker Compose v2+
• Go 1.25+ (for building from source)

Installation

# Install via script (recommended)
curl -fsSL https://raw.githubusercontent.com/agentregistry-dev/agentregistry/main/scripts/get-arctl | bash

# Or download binary directly from releases
# https://github.com/agentregistry-dev/agentregistry/releases

Start the Registry

# Start the registry server and look for available MCP servers
arctl mcp list

# The first time the CLI runs, it will automatically start the registry server daemon and import the built-in seed data.

Access the Web UI

To access the UI, open http://localhost:12121 in your browser.

📚 Core Concepts

MCP Servers

MCP (Model Context Protocol) servers provide tools, resources, and prompts to AI agents. They're the building blocks of agent capabilities.

Agent Gateway

The Agent Gateway is a reverse proxy that provides a single MCP endpoint for all deployed servers:

sequenceDiagram
    participant IDE as AI IDE/Client
    participant GW as Agent Gateway
    participant FS as filesystem MCP
    participant GH as github MCP
    
    IDE->>GW: Connect (MCP over HTTP)
    GW-->>IDE: Available tools from all servers
    
    IDE->>GW: Call read_file()
    GW->>FS: Forward to filesystem
    FS-->>GW: File contents
    GW-->>IDE: Return result
    
    IDE->>GW: Call create_issue()
    GW->>GH: Forward to github
    GH-->>GW: Issue created
    GW-->>IDE: Return result

IDE Configuration

Configure your AI-powered IDEs to use the Agent Gateway:

# Generate Claude Desktop config
arctl configure claude-desktop

# Generate Cursor config
arctl configure cursor

# Generate VS Code config
arctl configure vscode

🤝 Get Involved

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

Show your support

• 🐛 Report bugs and issues: GitHub Issues
• 💡 Suggest new features: GitHub Discussions
• 🔧 Submit pull requests: GitHub Repository
• ⭐ Star the repository: Show your support on GitHub
• 💬 Join the Conversation: Join our Discord Server

Related Projects

• Model Context Protocol
• kagent
• MCP Go SDK
• FastMCP

📚 Resources

• 📖 [Documentation] Coming Soon!
• 💬 GitHub Discussions
• 🐛 Issue Tracker

📄 License

MIT License - see LICENSE for details.