Agentregistry

Agentregistry is an open source, Kubernetes native registry for MCP servers, agents, skills and prompts. Platform teams use it to curate the AI building blocks their organization trusts in one catalog, then deploy them into a cluster through kagent with a single command. It is built for platform and DevOps teams that want governance over which AI components developers can use. MCP servers can be pulled from npm, PyPI, OCI or a remote URL, searched by name, publisher or tool, and deployed as Kubernetes workloads, with enrichment scores and verified publisher signals on every entry. Skills are versioned artifacts that can be scaffolded, published and pulled into a repository by name. Agent blueprints record an agent's identity, model, servers, skills and prompts, and kagent materializes them as native Kubernetes resources, with approval before deployment and support for many connected clusters. A gateway gives AI clients a single secured endpoint and routes each tool call to the right backend MCP server, and the arctl command line tool generates configuration for Claude Desktop, Cursor, Kiro and VS Code. Agentregistry is free under the Apache 2.0 license and installs into a cluster with a Helm chart. No hosted edition or paid plan is published. The product depends on Kubernetes and kagent, so it fits teams already running workloads on a cluster rather than individual developers working locally.