agentregistry-dev/agentregistry

Fast-track AI innovation with a centralized, trusted, curated registry

What it solves

Agentregistry provides a centralized platform to manage the fragmented ecosystem of AI building blocks. Instead of developers manually configuring MCP servers, agents, and skills scattered across various registries (npm, PyPI, Docker Hub), it offers a single, trusted source of truth for discovery, deployment, and governance within an organization.

How it works

It operates as a registry with a CLI (arctl) and a web UI that allows users to package and publish AI artifacts. It supports registering MCP servers from multiple sources (npm, PyPI, OCI/Docker, or HTTP/SSE), as well as creating versioned "skills" (knowledge packages) and "agents" (bundles of identity and dependencies). These artifacts can be deployed locally or to Kubernetes via Helm, and are integrated with AI clients like Claude Desktop, Cursor, and VS Code through a companion proxy called agentgateway.

Who it’s for

  • Platform Teams: To curate, approve, and standardize AI artifacts for company-wide use.
  • Developers: To quickly discover, build, and publish AI tools and agents with minimal manual configuration.

Highlights

  • Unified Catalog: A single place to find and manage MCP servers, agents, skills, and prompts.
  • Multi-Source Support: Registers artifacts from npm, PyPI, Docker, and remote endpoints.
  • IDE Integration: Automatically generates configurations for popular AI-enabled IDEs (Cursor, VS Code).
  • Governance Tools: Centralized control over versioning, metadata, and approved artifacts.
  • Kubernetes Ready: Supports consistent deployment workflows from local development to cluster environments.