Members from various AI protocols (MCP, A2A, and others) are collaborating on a common AI Catalog standard for discovering heterogeneous AI artifacts across the ecosystem.

Contact us via GitHub Discussions, Issues, and Pull Requests.

The AI Catalog specification is built from specification/ai-catalog.md and published to GitHub Pages by the workflow in .github/workflows/publish-spec.yml.

Published site: agent-card.github.io/ai-catalog

Pushes to the main branch update the canonical published site. Same-repo pull requests publish rendered preview pages under https://agent-card.github.io/ai-catalog/pr/<number>/, including a rendered diff preview against the PR base branch. The workflow also keeps a pull request comment updated with the live preview URL while the PR remains open.

GitHub Pages for the repository should be configured to serve from the gh-pages branch at the repository root.

To build the published HTML locally:

uv run --locked python tools/build_spec.py specification/ai-catalog.md dist/index.html --config specification/respec-config.json

This uses uv to resolve the Markdown dependency and writes the generated page to dist/index.html.

The build dependencies live in pyproject.toml and are pinned in uv.lock.

If you want to omit the explicit paths, the builder defaults to the spec source and dist/index.html output:

uv run --locked python tools/build_spec.py

If you change the spec build dependencies, refresh the lockfile with:

uv lock

The generated site entry point is dist/index.html.

There are multiple evolving standards for communication protocols between AI clients and servers. While these protocols differ in capabilities and technical architecture, there are common ecosystem needs for producers and consumers of services built on top of them. This includes discovery, verifiable metadata, and trusted identity standards.

Without a common discovery standard, we see duplicative and incompatible efforts in the ecosystem, such as registry, marketplace, identity providers, UI and payments extensions, and others. This increases complexity, risk of vulnerabilities, and harms interop within the developer community.

In this repo, we are defining the AI Catalog: a typed, nestable JSON container for discovering heterogeneous AI artifacts. Each catalog entry identifies an artifact by media type and may either reference the native artifact metadata by URL or inline it directly. The specification also defines an optional Trust Manifest extension for identity, attestations, and provenance metadata.

Important Note: The AI Catalog standard does not replace or redefine protocol-specific artifact formats. It provides a common discovery and trust layer around them.

The project defines a schema for a secure, extensible AI Catalog. This may include items such as:

• Common catalog metadata, such as publisher, documentation links, descriptions, icons, and versioning.
• Verifiable metadata, such as attestations, certifications, provenance, and trust signals.
• Identity metadata, such as Decentralized Identifiers (DID) or Secure Production Identity Framework For Everyone (SPIFFE).
• Custom metadata defined by individual producers or consumers for their own needs.

Actual properties for inclusion will be debated in PRs.

AI Catalog entries can reference or inline a wide range of AI artifacts, including MCP servers, A2A agent cards, Claude Code plugins, datasets, model cards, and nested catalogs.

Protocol-specific metadata remains defined by the underlying artifact specifications. The catalog does not duplicate or constrain those schemas; it uses media types to identify what each entry contains.

Nested AI Catalog entries allow publishers to organize large catalogs, delegate sub-catalogs, or package related artifacts together.

As part of the specification, this project defines a standard way to publish a catalog at a well-known URL for a given domain, for example at ./well-known/ai-catalog.json.

Implementing protocols may support dynamic card creation and discovery through their own methods for scenarios such as providing different card content based on a caller's identity.

Defining those protocol-specific dynamic behaviors is NOT in scope for the AI Catalog specification.

The AI Catalog project should also define a common AI Catalog Registry standard that provides a universal interface for clients to interact with catalog collections.

This is a temporary working repo maintained by the Linux Foundation. Proposals and changes will be made in public in this repo.

This project will be moved to a permanent location a later date with a permanent governance model.

When the specification is finalized, A2A and MCP steering committees will vote on adoption of the AI Catalog standard, potentially replacing existing protocol-specific standards or proposals.

Protocol-specific artifacts would remain in their native formats and become discoverable through common AI Catalog entries.

If adopted, MCP and A2A steering committees will recommend duplicative card-adjacent efforts be consolidated, such as Registry, Agent Identity, and others.