capa

Agentic Capabilities and Package Manager

CAPA is the package manager for AI coding agents. Declare your skills, tools, rules, sub-agents, MCP servers, and plugins once in capabilities.yaml, run capa install, and CAPA writes them into Cursor, Claude Code, Codex, Windsurf, GitHub Copilot, and 35+ other agents.

https://github.com/user-attachments/assets/98442d19-44c9-43e6-b2c2-88156b189d5e

Why CAPA?

AI coding agents need rules, tools, and conventions — and right now those live scattered across CLAUDE.md, .cursor/rules/, AGENTS.md, MCP configs, and a half-finished internal doc. No two setups match.

CAPA collapses it into one capabilities.yaml next to your code: skills, tools, rules, sub-agents, plugins. capa install fans it out to every provider in its native format — .cursor/rules/ for Cursor, .claude/agents/ and CLAUDE.md for Claude Code, AGENTS.md for Codex, and so on. Capa-managed marker blocks keep your hand-written content untouched.

One file, version controlled, pinned by capabilities.lock, cached by SHA. The teammate who clones tomorrow gets the exact bytes you got today.

What it does

• One capabilities.yaml manages the content for your agent. Write rules, hooks, and tools once — capa runs them natively, supporting 35+ agents (Cursor, Claude Code, Codex, Windsurf, Copilot, Gemini CLI, and more). No more keeping .cursor/rules/, .claude/settings.json, and AGENTS.md in sync by hand.
• 19–40% cheaper inference, same quality. One MCP server per agent, tools lazy-load on demand instead of front-loading the whole catalog. Measured across 150 trials on claude-opus-4-8.
• Sub-agents only see the tools they need. Each gets its own filtered MCP endpoint — so your research sub-agent isn't holding a git push tool it shouldn't touch.

Installation

macOS and Linux:

curl -LsSf https://capa.infragate.ai/install.sh | sh

Windows:

powershell -ExecutionPolicy ByPass -c "irm https://capa.infragate.ai/install.ps1 | iex"

Quick start

1. Initialize your project

cd your-project
capa init

This creates a capabilities.yaml next to your code.

2. Install

capa install

3. Boostrap (Optional)

If you are already working on a project, and would like to get all of it's capabilities be managed by CAPA, simply use the /bootstrap skill. The agent will scan your project for skills, MCPs, relevant tools, hooks, and rules.

capa install resolves SHAs and downloads anything that isn't already in the cache. It then writes the per-provider files (.cursor/rules/, .claude/agents/, AGENTS.md, and so on) and registers one MCP endpoint with each agent on your list. The resolved SHAs land in capabilities.lock so the next clone gets the same bytes.

4. Use capa sh

capa sh                                  # list every configured tool
capa sh brave                            # list brave subcommands
capa sh brave search --query "…"         # run a tool directly

Every tool you define is also a CLI command under capa sh. MCP tools live at capa sh <server> <tool>. Shell tools live at the top level (or under whatever group you assigned).

Documentation

Guides, the full schema reference, and the registry catalog:

https://capa.infragate.ai

License

MIT