Observal
A self-hosted registry and control plane for an organization's internal Skills, MCP servers and Agents, adding a discoverability layer and usage insights so teams reuse components instead of rebuilding them.
Observal is a control plane and system of record for an organization's internal AI components: Skills, Agents, MCP servers and similar building blocks that teams create to boost productivity but often fail to reuse. It addresses two problems it identifies in most organizations: components stored in siloed, poorly documented repositories that make similar existing work hard to find, and a missing feedback loop, since AI tools tend to fail silently with subtly wrong answers instead of clear error codes, leaving little visibility into how a component is actually used.
It provides a centralized discovery layer for these components alongside insight into how they are used, aiming to turn silent failures into actionable feedback so internal AI tools are continuously improved for the people using them. It supports Claude Code, Cursor, Kiro, Pi, Copilot, Codex, OpenCode and other agent tools, and lets a team package Skills, MCP servers, hooks, prompts and sandboxes into one versioned, reusable agent, with a governed registry for reviewing and approving submissions.
It is written in Python and licensed under Apache-2.0.