THE WORKSHOP / DEVELOPER TOOLS / 01

AI API endpoints.
One less thing to look up.

Find a provider. Copy the base URL. Get back to building.

224 entries · 12 documentation-checked setups · Updated 2026-09-23

224 providers & endpoint variants

Download JSON ↓

“Docs checked” means compared with official documentation, not a live API test. Other entries come from the Models.dev community catalogue. About the data ↓

BEFORE YOU CONNECT

The right URL, in the right place.

Base URL ≠ request endpoint

An SDK adds a route to the base URL. For example, Chat Completions commonly adds /chat/completions. Pasting the full request URL into base_url can duplicate the path.

Compatibility has limits

An OpenAI-compatible interface does not guarantee support for every model, parameter or tool. Use the destination provider’s key and exact model ID. Check its documentation before switching a production integration.

Keys stay in your project

No signup or API key is needed to use this directory. Examples run on your server or local machine. Set credentials as environment variables, and choose a model your account can access.

A useful snapshot, not every API

This directory covers model providers, gateways, regional plans and local runtimes. Community metadata comes from Models.dev ↗, under its MIT licence. It is not a ranking, endorsement or live uptime monitor.

How do I fix a 401 or 404?

For a 401, check your key, account permissions and region. For a 404, check the model ID and whether your SDK duplicated a route. Some providers need a workspace, deployment or account ID in the URL.

Found an outdated endpoint? Send a correction with the provider’s documentation ↗.

Building a workflow around these APIs? Explore our content automation work →