Friz-zy/ai-capability-registry

AI Capability Registry

Warning Connecting this registry through AGENTS.full-registry.md.template makes agents read routing indexes and selected capability guides at runtime. This can use substantially more model context and tokens than a minimal static agent setup. Prefer task-, role-, keyword-, or project-specific subsets when context budget matters.

AI Capability Registry is an experiment in dynamic capability routing for AI agents: load the right skills, MCP servers, workflows, and integration instructions only when the current task needs them.

Instead of stuffing every agent with a huge static prompt, a random pile of tools, or every available MCP server, this repository treats agent capabilities as versioned infrastructure. An agent can resolve the user's request by task, role, and keywords, then progressively load only the relevant skill instructions or MCP connection guidance.

This is not the theoretical optimum for every setup. A carefully hand-tuned agent with the smallest possible set of skills and MCP servers for one specific task will usually be simpler, faster, and easier to reason about. The registry is meant for the more common middle ground: teams and users who run many agents across many tasks and need something better than enabling a broad list of skills and MCP servers for every agent "just in case". Many agent tools make that broad attachment path easy; this project explores a more explicit routing layer between fully minimal per-agent configuration and everything-enabled-by-default capability sprawl.

You can also use the registry as a starting point for minimal static setups. A user can ask their AI assistant to inspect the shared registry and assemble the smallest useful set of skills and MCP servers for a specific role, task, or agent configuration. In that mode, the registry is not the runtime layer itself; it is the common source of reviewed capabilities from which task-specific agent setups are derived.

The goal is to make