Overview
One OpenAI-compatible endpoint for publicly available AI capacity.
llmfaucet is a public, OpenAI-compatible gateway for officially available anonymous/keyless AI capacity. It normalizes requests, selects a compatible healthy route, and exposes standard response and rate-limit signals.
This is a shared experimental service. Provider capacity, model availability, limits, and latency can change. It is not an SLA or a promise of a particular model.
Start here
| Goal | Guide |
|---|---|
| Send your first request | Quickstart |
| Configure a coding tool | Coding agents |
| Understand fair-use limits | Limits and fair use |
| Inspect provider health | Status page |
| Join controlled preview testing | Early testers |
| Open the deployed preview | Preview environment |
| Understand data handling | Privacy |
Endpoint contract
Production and preview are isolated environments. Use the URL supplied for the environment you are testing:
Production: https://api.llmfaucet.dev/v1
Preview: https://api.llmfaucet-preview.dev/v1If those domains are not configured yet, use the Worker URL printed by Wrangler
and keep /v1 as the API base path.
Model selectors
auto General automatic routing
auto:fast Prefer lower-latency healthy routes
auto:smart Prefer higher-ranked available routes
auto:coding Prefer coding-tagged routesSelectors are routing preferences, not guarantees. The selected route is
reported in X-Routed-Via when available.
Compatibility surfaces
- OpenAI chat completions:
POST /v1/chat/completions - OpenAI legacy completions:
POST /v1/completions - OpenAI models:
GET /v1/models - Anthropic messages:
POST /v1/messages - Responses compatibility:
POST /v1/responses - Embeddings:
POST /v1/embeddingswhen an embedding route is available - Service health:
GET /healthandGET /status
Read the quickstart for working requests and the limits guide before connecting an unattended client.