Inference GatewayRegistry

Discover A2A (Agent-to-Agent) services and portable skills you can drop into any agent.

🤖

ADL-compliant agents

Every listed agent ships an `agent.yaml` that follows the Agent Definition Language schema, so capabilities, tools, and skills are uniformly typed and discoverable.

🧩

Portable skills

Drop-in markdown playbooks installable with `infer skills install `. Vendor-neutral and reusable across any ADL agent.

🛰️

Live catalog

Agent and skill entries are fetched at runtime from the catalog repos via jsDelivr. Listing a new entry takes a PR - no redeploy of this site.

🚀

One-command install

Each card surfaces a ready-to-paste `infer agents add` or `infer skills install` command and, where available, the OCI image tag.

🔌

Vendor-neutral

Swap providers (OpenAI, Anthropic, DeepSeek, Google, Mistral, Ollama, Groq) and runtimes without rewriting the agent. ADL keeps everything portable.

🛠️

How-to guides

Step-by-step instructions for browsing the catalog, installing entries with the `infer` CLI, and contributing your own agent or skill.

Submit yours

The registry indexes two external catalogs - it does not store agent or skill manifests itself.

Agents live in inference-gateway/agents. Any public GitHub repo that ships an ADL agent.yaml at its root is eligible. Open a PR adding one entry to agents.yaml with the repo URL and an optional ref.
Skills live in inference-gateway/skills. PRs add per-skill files. Each skill carries an SPDX license field.

CI in each catalog repo validates the entry and regenerates catalog.json on merge; the new entry appears here within the jsDelivr @main cache window (up to ~12h).