ara-registry

AI Registry for Agents (ARA)

ARA (pronounced ah-rah) is an open package registry and distribution system for AI development artifacts. It provides a standardized way to discover, share, and install reusable components that enhance AI assistants and development workflows.

Think npm for AI — publish once, use everywhere.

What can you find in ARA?

Package Type Description
Custom Agents Agent configurations with prompts, tools, and behaviors
MCP Servers Model Context Protocol servers that extend AI capabilities
Context Knowledge files, prompt templates, and reference materials
Skills Procedural knowledge via SKILL.md that agents load dynamically
Powers Bundles of MCP tools, steering files, and hooks for specialized knowledge
Steering Persistent knowledge about your projects
AGENTS.md An open format for guiding coding agents

Platform Agnostic

ARA is designed to work across AI coding agents and agentic frameworks — Kiro, Claude Code, Strand, and others. The specification defines artifacts that are portable, not locked to a single tool or framework.

How It Works

ARA is built on two open specifications:

  • ara.json — A standardized package manifest format that anyone can use to create compatible packages
  • Registry API — An HTTP API specification that allows anyone to build compatible registries

ARA operates as a hybrid registry: it can host packages directly or act as a metaregistry pointing to existing sources like npm, PyPI, Docker, and git.

Repositories

Repository Description
spec ARA specification, ara.json format, registry API, and JSON schema

Tenets

  • Community-Driven Discovery — Usage patterns and community feedback surface valuable artifacts, not editorial gatekeeping
  • Platform Agnostic — Artifacts work across AI coding agents, not just one platform
  • Convention Over Configuration — Sensible defaults reduce friction for authors and users
  • Transparent Governance — Decisions are made openly with community input; no single organization controls the spec

Get Involved

We welcome contributions from the community. Check out the spec repository to get started, or open an issue to share your ideas.

License

This project is licensed under the Apache-2.0 License.