llmfaucet

One endpoint. Free AI models. No keys, no setup.

A public, OpenAI-compatible gateway that routes requests across AI providers with officially available anonymous access.

Website · API Status · Documentation · Sponsor

Warning

llmfaucet is built for experimentation, learning, prototypes, open-source projects, and personal workflows. Free upstream capacity can change, slow down, or disappear without notice. Do not depend on it for critical production workloads.

Why llmfaucet?

Most AI providers have free models, trial capacity, or anonymous endpoints. Using them separately is frustrating: every service has a different API format, model list, rate limit, and availability pattern.

llmfaucet gives AI agents and applications one public endpoint instead.

Your agent or application
          │
          │  https://api.llmfaucet.dev/v1
          ▼
      llmfaucet
          │
          ├── Pollinations
          ├── LLM7
          ├── OpenCode Zen
          ├── OVH AI Endpoints
          └── AI Horde

The gateway selects a compatible healthy model, retries another source when one is unavailable, and returns an OpenAI-compatible response.

Features

Feature	Description
Public endpoint	Use one hosted API URL without self-hosting infrastructure
No upstream keys	llmfaucet only integrates sources with anonymous or keyless access
OpenAI compatible	Works with clients that support `/v1/chat/completions`
Agent-friendly	Designed for Claude Code, Codex CLI, Cline, Roo Code, Continue, Aider, and other coding tools
Smart routing	Selects models based on capability, health, speed, and current availability
Automatic fallback	Tries another compatible upstream after rate limits, upstream failures, or timeouts
Streaming	Supports streamed responses where the selected upstream supports streaming
Model selectors	Choose `auto`, `auto:fast`, `auto:smart`, or `auto:coding`
Honest limits	Returns rate-limit and retry headers so tools can back off correctly
Privacy-first	No upstream API keys are collected from free users

Quick start

Use the public OpenAI-compatible endpoint:

curl https://api.llmfaucet.dev/v1/chat/completions \
  -H "Authorization: Bearer free" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [
      {
        "role": "user",
        "content": "Explain what an API gateway does in one paragraph."
      }
    ]
  }'

The Authorization header is optional for anonymous use in most supported clients, but some SDKs require a value. Use free when a placeholder API key is required.

Model selectors

You do not need to know which upstream model is available. Use a selector and let llmfaucet choose.

Model	Best for	Routing behavior
`auto`	General use	Balanced quality, speed, and availability
`auto:fast`	Quick tasks	Prioritizes lower latency models
`auto:smart`	Harder reasoning	Prioritizes stronger available models
`auto:coding`	Coding agents	Prioritizes code-focused models and tool-capable routes
Specific model ID	Advanced users	Requests a named model when it is currently available

Example:

{
  "model": "auto:coding",
  "messages": [
    {
      "role": "user",
      "content": "Write a TypeScript function that retries fetch requests with exponential backoff."
    }
  ]
}

Supported capacity sources

llmfaucet only uses publicly available, anonymous, or explicitly keyless AI sources. Availability changes often, so the live catalog is always the source of truth.

Source	Role in llmfaucet	Typical capabilities
Pollinations	Primary general-purpose capacity	Chat, code, image, audio, and selected multimodal models
LLM7	General and reasoning fallback	General chat, coding, reasoning, long-form tasks
OpenCode Zen	Coding-focused capacity	Code generation, agent tasks, software engineering
OVH AI Endpoints	Reliable open-model fallback	Open-weight text models, selected vision and media models
AI Horde	Community overflow capacity	Community-hosted text and image models

Catalog expectations

Models may be added, removed, renamed, or temporarily unavailable.
A listed upstream may reject a request due to its own policies or limits.
The same model can exist through multiple upstreams.
llmfaucet may route a request to a compatible alternative when the requested model is unavailable.
The response includes routing headers so you can inspect what served the request.

X-Routed-Via: pollinations/qwen3-coder
X-Fallback-Attempts: 1
X-RateLimit-Remaining: 17
X-RateLimit-Reset: 2026-08-25T00:00:00Z

How routing works

Every request goes through a small routing process.

llmfaucet checks whether the request needs special support such as images, tools, structured JSON, or long context.
It removes models that cannot support those requirements.
It removes sources currently marked unhealthy, overloaded, or rate-limited.
It scores the remaining routes using model capability, observed speed, and recent availability.
It forwards the request to the best route.
If an upstream returns a temporary error, llmfaucet retries another compatible route.
If all compatible sources are busy, the request is queued or receives a retry response.

Request
  │
  ├── Need vision, tools, JSON, or code?
  │
  ├── Remove incompatible models
  │
  ├── Remove unhealthy or rate-limited sources
  │
  ├── Rank by selector, speed, quality, and load
  │
  ├── Route to best available source
  │
  └── Retry compatible fallback when needed

Usage limits

Free AI capacity is limited. llmfaucet shares that capacity fairly so one user cannot consume the entire service.

Access level	Daily request limit	Queue priority	How to access
Anonymous	20 requests	Standard	Use the endpoint with `Bearer free`
Registered	50 requests	Higher	Create a free llmfaucet key
GitHub Sponsor	Tier-based	Priority	Sponsor the project and link GitHub

Limits are enforced per IP address for anonymous users and per llmfaucet API key for registered users. Limits reset daily at midnight UTC.

Tip

Use auto:fast for short tasks, keep prompts focused, and avoid unnecessary retries. This helps free capacity serve more people.

GitHub Sponsors

llmfaucet is sustained by the people who use it. GitHub Sponsors supports the open-source project while unlocking higher limits and priority access.

Tier	Monthly support	Daily request limit	Benefits
Free	$0	20	Public endpoint and standard queue access
Registered	$0	50	Free API key and higher queue priority
Supporter	$3/month	200	Priority queue and higher daily capacity
Pro	$5/month	500	Higher priority and access to early features
Builder	$10/month	1,000	Highest priority and experimental model access

After sponsoring:

Return to llmfaucet.dev.
Sign in with GitHub.
Link your active GitHub Sponsors account.
Generate an llmfaucet-... API key.
Paste the key into your AI coding tool or application.

If a sponsorship ends, the key remains valid and automatically falls back to the free tier instead of breaking your existing setup.

Become a GitHub Sponsor →

Coding agents

llmfaucet is designed to work with common coding agents and OpenAI-compatible tools.

Tool	Provider type	Base URL	Model
Cline	OpenAI Compatible	`https://api.llmfaucet.dev/v1`	`auto:coding`
Roo Code	OpenAI Compatible	`https://api.llmfaucet.dev/v1`	`auto:coding`
Continue	OpenAI	`https://api.llmfaucet.dev/v1`	`auto`
Aider	OpenAI Compatible	`https://api.llmfaucet.dev/v1`	`openai/auto:coding`
Codex CLI	Custom provider	`https://api.llmfaucet.dev/v1`	`auto:coding`
Claude Code	Anthropic Compatible	`https://api.llmfaucet.dev`	`auto:coding`
Cursor	Custom OpenAI API	`https://api.llmfaucet.dev/v1`	`auto:coding`

Cline and Roo Code

Open provider settings.
Select OpenAI Compatible.
Set the base URL to:

https://api.llmfaucet.dev/v1

Set API key to free, or use your llmfaucet-... key.
Select auto:coding.

Aider

aider \
  --openai-api-base https://api.llmfaucet.dev/v1 \
  --openai-api-key free \
  --model openai/auto:coding

Replace free with your llmfaucet API key to use your registered or sponsor allowance.

Claude Code

export ANTHROPIC_BASE_URL="https://api.llmfaucet.dev"
export ANTHROPIC_AUTH_TOKEN="free"

claude

Note

Use ANTHROPIC_AUTH_TOKEN, not ANTHROPIC_API_KEY. Restart Claude Code after changing these variables.

Codex CLI

Add the following to ~/.codex/config.toml:

[model_providers.llmfaucet]
name = "llmfaucet"
base_url = "https://api.llmfaucet.dev/v1"
env_key = "LLMFAUCET_API_KEY"
wire_api = "responses"

Then set your key:

export LLMFAUCET_API_KEY="free"

Use your llmfaucet-... key instead of free for registered or sponsor access.

API compatibility

Endpoint	Status	Purpose
`GET /v1/models`	Planned	List currently available models and capabilities
`POST /v1/chat/completions`	In progress	OpenAI-compatible chat completions
`POST /v1/completions`	Planned	Legacy prompt completion and editor autocomplete
`POST /v1/responses`	Planned	OpenAI Responses API support for Codex CLI
`POST /v1/messages`	Planned	Anthropic Messages API support for Claude Code
`POST /v1/messages/count_tokens`	Planned	Anthropic-compatible token counting
`POST /v1/embeddings`	Planned	Compatible embedding routing by model family
`GET /status`	Planned	Public upstream health and service status

Architecture

llmfaucet stays lightweight by acting as a routing layer instead of running expensive models itself.

Component	Purpose
Cloudflare Workers	Public API gateway, request validation, routing, streaming, and fallback
Cloudflare KV	Daily budgets, rate-limit counters, cooldowns, and active catalog cache
Cloudflare D1	Catalog history, aggregate statistics, and operational records
Cloudflare Queues	Fair waiting room when free sources are saturated
Cloudflare Pages	Documentation, landing page, setup guides, and public status site
Oracle Cloud Always Free	Scheduled upstream probes, model health checks, and catalog updates

Coding agent or application
          │
          ▼
api.llmfaucet.dev
          │
          ▼
Cloudflare Worker
  ├── Budget and abuse checks
  ├── Capability filtering
  ├── Health-aware routing
  ├── Short fallback chain
  └── Stream relay
          │
          ▼
Anonymous-capacity providers

Development

Branch strategy

Branch	Environment	Domains
`main`	Production	`llmfaucet.dev` and `api.llmfaucet.dev`
`dev`	Preview / staging	`llmfaucet-preview.dev` and `api.llmfaucet-preview.dev`

Production and preview use isolated Cloudflare Workers, KV namespaces, D1 databases, and queue resources. Preview traffic never consumes production budgets or changes production routing state.

Local setup

git clone https://github.com/Hyperkit-dev/llmfaucet.git
cd llmfaucet
npm install
npm run dev

Create a local environment file:

cp .dev.vars.example .dev.vars

Run checks before opening a pull request:

npm run lint
npm run test
npm run typecheck

Roadmap

Trust and safety

llmfaucet follows a few non-negotiable rules:

Only use providers that document anonymous, keyless, or otherwise permitted public access.
Do not scrape private services, bypass paywalls, rotate accounts, or automate account creation.
Do not expose upstream provider credentials because llmfaucet does not collect them for public routing.
Apply fair-use limits to prevent one user from exhausting shared capacity.
Publish service status and known upstream limitations honestly.
Treat prompts as operational data only; minimize retention and do not sell user prompts.

Contributing

Contributions are welcome.

Useful areas include:

New officially permitted anonymous providers
Provider health checks and capability detection
OpenAI, Anthropic, and Responses API compatibility
Coding-agent setup guides
Rate-limit and routing improvements
Documentation, examples, and integrations
Security review and abuse prevention

Before opening a pull request, please include tests and clearly document upstream access terms for any provider integration.

Read the contributing guide: CONTRIBUTING.md

Support

If llmfaucet helps your work, consider supporting its maintenance through GitHub Sponsors.

License

llmfaucet is released under the MIT License.

The project may include adapted ideas or attributed implementation patterns from compatible open-source projects. See NOTICE.md for third-party acknowledgements.

Name		Name	Last commit message	Last commit date
.github		.github
bin		bin
docs		docs
public		public
scripts		scripts
src		src
test		test
.env.example		.env.example
.gitignore		.gitignore
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
SECURITY.md		SECURITY.md
ecosystem.config.cjs		ecosystem.config.cjs
package-lock.json		package-lock.json
package.json		package.json
schema.sql		schema.sql
tsconfig.json		tsconfig.json
wrangler.toml		wrangler.toml

Navigation Menu

Repository files navigation

llmfaucet

One endpoint. Free AI models. No keys, no setup.

Why llmfaucet?

Features

Quick start

Model selectors

Supported capacity sources

Catalog expectations

How routing works

Usage limits

GitHub Sponsors

Coding agents

Cline and Roo Code

Aider

Claude Code

Codex CLI

API compatibility

Architecture

Development

Branch strategy

Local setup

Roadmap

Trust and safety

Contributing

Support

License

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

llmfaucet

One endpoint. Free AI models. No keys, no setup.

Why llmfaucet?

Features

Quick start

Model selectors

Supported capacity sources

Catalog expectations

How routing works

Usage limits

GitHub Sponsors

Coding agents

Cline and Roo Code

Aider

Claude Code

Codex CLI

API compatibility

Architecture

Development

Branch strategy

Local setup

Roadmap

Trust and safety

Contributing

Support

License

About

Resources

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages