Content
<p align="center">
<picture>
<source media="(prefers-color-scheme: dark)" srcset="docs/img/agentregistry-logo.png">
<source media="(prefers-color-scheme: light)" srcset="docs/img/agentgateway-logo-light.png">
<img src="docs/img/agentregistry-logo.png" alt="agentregistry" width="500"/>
</picture>
</p>
<h1 align="center" style="font-size: 3em;">Build. Deploy. Discover.</h1>
<h3 align="center">One registry for MCP servers, agents, skills and prompts.</h3>
<br/>
<p align="center">
<a href="https://github.com/agentregistry-dev/agentregistry/stargazers"><img src="https://img.shields.io/github/stars/agentregistry-dev/agentregistry?style=social" alt="GitHub Stars"></a>
<a href="https://discord.gg/HTYNjF2y2t"><img src="https://img.shields.io/discord/1435836734666707190?label=Discord&logo=discord&logoColor=white&color=5865F2" alt="Discord"></a>
<a href="https://github.com/agentregistry-dev/agentregistry/releases"><img src="https://img.shields.io/github/v/release/agentregistry-dev/agentregistry?label=Release" alt="Release"></a>
<a href="LICENSE"><img src="https://img.shields.io/badge/License-Apache%202.0-green.svg" alt="License"></a>
<a href="https://golang.org/doc/install"><img src="https://img.shields.io/badge/Go-1.25+-blue.svg" alt="Go Version"></a>
</p>
<p align="center">
<a href="https://aregistry.ai">Website</a> · <a href="https://aregistry.ai/docs/">Docs</a> · <a href="https://discord.gg/HTYNjF2y2t">Discord</a> · <a href="https://github.com/agentregistry-dev/agentregistry">GitHub</a> · <a href="#quick-start">Quick Start</a>
</p>
---
## What is agentregistry?
agentregistry is an open-source platform that gives you one place to find, manage, and run MCP servers, AI agents, and skills.
Right now, the MCP servers and AI tools your team needs are spread across npm, PyPI, Docker Hub, GitHub repos, and random URLs. Nobody knows which ones are trustworthy, which versions work, or how to get them running. Every developer is doing their own manual Docker setup and IDE configuration.
agentregistry puts all of that into a single registry with a CLI and a web UI. You publish artifacts once, and then anyone on your team can discover them, deploy them with one command, and have their IDE automatically configured to use them.
---
## Why agentregistry?
- **One trusted source for AI building blocks** — a curated catalog instead of scattered repos, scripts, and one-off MCP setup
- **Faster developer onboarding** — discover approved artifacts quickly with less manual configuration
- **Consistent path from laptop to cluster** — same discovery and delivery workflow across local dev and Kubernetes
- **Governance without slowing teams down** — centralize curation and publishing without forcing each team to rebuild the process
<p align="center">
<img src="docs/img/hero-final.png" alt="agentregistry overview" width="800"/>
</p>
<table>
<tr>
<td width="50%" valign="top">
<h3>For Organizations</h3>
<p><strong>Curate & Deploy</strong></p>
<p>Package, publish, and curate AI artifacts in a single centralized registry.</p>
<ul>
<li><strong>Centralized Control</strong> - Package and publish AI artifacts into a single registry</li>
<li><strong>Security & Governance</strong> - Curate and approve agents, servers, and skills before company-wide deployment</li>
<li><strong>Resource Metadata</strong> - Add context to help teams assess trustworthiness and security</li>
</ul>
</td>
<td width="50%" valign="top">
<h3>For Developers</h3>
<p><strong>Build & Publish</strong></p>
<p>Build, test, publish, and deploy AI artifacts with minimal dependencies.</p>
<ul>
<li><strong>Local Development</strong> - Create and test agents, skills, and MCP servers locally</li>
<li><strong>Easy Publishing</strong> - Publish your artifacts to a registry with a single command</li>
<li><strong>Pull & Run Anywhere</strong> - Pull artifacts from the registry and run them in any environment instantly</li>
<li><strong>Discover & Consume</strong> - Find new artifacts to add to registry or optimize existing artifacts</li>
</ul>
</td>
</tr>
</table>
<a id="quick-start"></a>
## Quick Start
**Prerequisites:** Docker Desktop with Docker Compose v2+
```bash
# 1. Install the CLI
curl -fsSL https://raw.githubusercontent.com/agentregistry-dev/agentregistry/main/scripts/get-arctl | bash
# 2. Download the Docker Compose file matching the installed CLI release
export VERSION="$(arctl version | awk 'NR == 1 { print $3 }')"
curl -fsSLo agentregistry-compose.yml \
"https://raw.githubusercontent.com/agentregistry-dev/agentregistry/${VERSION}/docker/docker-compose.yml"
# 3. Start AgentRegistry and its bundled PostgreSQL database
docker compose -f agentregistry-compose.yml up -d --wait
```
Open the AgentRegistry UI at http://localhost:12121. To stop the registry while
preserving its data, run:
```bash
docker compose -f agentregistry-compose.yml down
```
---
## Core Capabilities
### Build
Create, scaffold, and publish the building blocks of your agentic infrastructure.
- **MCP servers** -- Register servers from npm (`npx`), PyPI (`uvx`), OCI/Docker images, or remote HTTP/SSE endpoints. Each entry supports versioning, environment variables, and package references.
- **Skills** -- Build structured knowledge packages that extend what an agent knows. A skill is a `SKILL.md` bundled with code examples, docs, PDFs, and reference URLs. Scaffold with `arctl init skill`, package and push the image with `arctl build ./skill --push`, then register the skill record with `arctl apply -f skill.yaml`.
- **Agents** -- Define agents that bundle an identity with dependencies: which MCP servers it needs, which skills it uses, and how it should be configured. Scaffold with `arctl init agent`, build and push the image with `arctl build ./agent --push`, then register the versioned agent record with `arctl apply -f agent.yaml`.
- **Prompts** -- Create reusable instruction templates that define how an agent should behave in specific contexts. Version and store them alongside agents, skills, and servers so they're discoverable and shareable across your team.
### Web UI
A browser-based admin interface at `localhost:12121`. Browse the artifact catalog, add MCP servers, skills, and agents, review metadata, manage deployments, and configure the registry -- all without touching the CLI.
<p align="center">
<img src="docs/img/arwui.gif" alt="agentregistry Web UI" width="800"/>
</p>
### Registry
Curate a shared catalog of MCP servers, agents, skills, and prompts your teams can trust and reuse.
- Publish artifacts to a central registry from npm, PyPI, Docker, OCI, or remote endpoints
- Discover approved artifacts through the CLI, REST API, or web UI at `localhost:12121`
- Give teams a consistent source of truth across environments
### Curation and Governance
Turn a broad set of available AI artifacts into a collection your organization is willing to support.
- Organize what developers can discover and deploy
- Review versioning, metadata, and environment variable requirements
- Standardize how artifacts are shared across teams
- Keep control of what gets published and promoted
### Deployment Workflows
Move from discovery to usage without reinventing the same delivery path for every team.
- Run workflows locally with `arctl`
- Deploy Agent Registry into Kubernetes with Helm
- Support local `arctl run` workflows and shared Kubernetes environments from the same registry
- Build and push agents — blueprints bundle an agent with its MCP servers and skills into a single deployable unit
### Client and Gateway Integration
Make approved artifacts easier to consume from the tools developers already use.
- Generate configuration for Claude Desktop, Cursor, Kiro IDE, and VS Code
- Pair with agentgateway for a consistent access layer to deployed MCP infrastructure
- Reduce manual setup for AI clients and shared environments
### How It Works Together
1. Platform teams curate and publish approved MCP servers, agents, and skills in Agent Registry
2. Developers discover those artifacts through the web UI or `arctl`
3. Teams run artifacts locally with `arctl` or deploy them to Kubernetes
4. AI clients and shared gateway infrastructure connect to approved artifacts through a consistent workflow
## Secure Access with agentgateway
agentregistry pairs with [agentgateway](https://github.com/agentgateway/agentgateway) to give you a single, secure entry point to all your deployed MCP servers and agents.
Instead of exposing every MCP server individually, agentgateway acts as an AI-native reverse proxy that sits in front of your entire agentic infrastructure:
- **Single endpoint** — AI clients (Claude Desktop, Cursor, VS Code) connect to one URL. The gateway routes each tool call to the correct backend MCP server.
- **Authentication & authorization** — Enforce identity and access policies before requests reach your MCP servers. Control who can call which tools. Supports JWT validation and on-behalf-of auth flows.
- **Centralized observability** — Log and monitor all agent-to-tool traffic in one place instead of instrumenting each server separately. Supports OTEL endpoints for traces, metrics, and logs.
- **Dynamic discovery** — Deploy a new MCP server through agentregistry and every connected client picks it up automatically — no reconfiguration needed.
- **LLM gateway** — agentgateway also acts as a unified gateway for LLM providers, giving you a single endpoint to route, manage, and secure access to multiple language models.
- **Transport flexibility** — Proxy across stdio, SSE, and streamable HTTP transports seamlessly.
<p align="center">
<img src="docs/img/arch-diagram.png" alt="agentgateway architecture diagram" width="800"/>
</p>
When you run `arctl apply -f deployment.yaml`, agentregistry automatically configures the gateway routing so your MCP servers are reachable through the secured proxy. Run `arctl configure cursor` to point your IDE at the gateway endpoint.
---
## Related Projects
| Project | Role |
|---|---|
| [agentgateway](https://github.com/agentgateway/agentgateway) | AI-native reverse proxy for MCP traffic |
| [kagent](https://github.com/kagent-dev/kagent) | Kubernetes-native AI agent platform |
| [kgateway](https://github.com/kgateway-dev/kgateway) | Cloud-native API gateway (Envoy + Gateway API) |
| [MCP Go SDK](https://github.com/modelcontextprotocol/go-sdk) | Go SDK for building MCP servers |
| [Model Context Protocol](https://modelcontextprotocol.io/) | The open standard for AI-to-tool communication |
---
## Community
### Communication channels
If you're interested in participating with the agentregistry community, come talk to us!
- We are available on [**Discord**](https://discord.gg/HTYNjF2y2t)
- To report security issues, please follow our [**vulnerability disclosure best practices**](https://github.com/agentregistry-dev/agentregistry/security)
- Find more information on the [**agentregistry website**](https://aregistry.ai)
### Community meetings
We do not yet have community meetings. [**Establishing these meetings**](https://github.com/agentregistry-dev/agentregistry/issues) is on our [**roadmap**](https://github.com/agentregistry-dev/agentregistry/issues). Please help us deliver this work by either commenting on the issue, or volunteering to establish the meetings.
### Contributing
See [`CONTRIBUTING.md`](CONTRIBUTING.md) for guidelines and [`DEVELOPMENT.md`](DEVELOPMENT.md) for architecture and local development setup.
[Report a bug](https://github.com/agentregistry-dev/agentregistry/issues) · [Suggest a feature](https://github.com/agentregistry-dev/agentregistry/discussions) · [Join Discord](https://discord.gg/HTYNjF2y2t)
## License
Apache 2.0 — see [`LICENSE`](LICENSE).
