01 / CONNECT
Hundreds of models.
One connection.
Use the familiar Claude Code interface with models from different providers through one API.
Open code. Open possibilities.
More models.
More models.
Fewer
limits.
Your Claude Code, with hundreds of free models. One router, dozens of providers, and automatic switching when a limit is reached.
Get the app
❯
npx freellmapi launch SMART ROUTINGDEMONSTRATION
groqInference
OpenRouter
▧Cerebras
NVIDIA
✦Gemini
MMistral
FreeLLMAPIONE ENDPOINT. EVERY MODEL.
request response
Llama 3.3 70BAUTO
Request routed through Groq
⌞YOU CREATE. WE FIND THE ROUTE.⌟
ONE ROUTER. AN ENTIRE ECOSYSTEM.And dozens of other providers
✦ Googlegroq▧ cerebras◉ NVIDIA⇶ OpenRouter
LESS FRICTION. MORE FLOW.
Stay in your flow.
Write code, try something new, think bigger. The router handles the limits so you can keep creating.
AUTOMATIC FAILOVERThe limit is reached.
The limit is reached.
Your ideas keep going.
A model returned 429 or a provider is unavailable? FreeLLMAPI automatically tries the next route. You keep working.
How routing worksrouting.logEXAMPLE
⌘Claude Codeyour request
01Llama 3.3 Groq429 limit
switching the route
02Qwen3 OpenRouter200 OK ✓
The request continues its journey
Hundreds of models.
02 / YOUR CONTROL
Your keys.
Under your control.
A local-first approach. The launch command passes the key to the Claude Code process without saving it in its configuration.
FIND YOUR MODEL
Different tasks. Your choice.
≈Reasoning
DeepSeek R1
✓Complex tasks, deep analysis, and code that takes a little thinking.
Free routes
✦General purpose
Gemini 2.5 Flash
✓A fast assistant for ideas, everyday tasks, and large contexts.
Free routes
Code & chat
Llama 3.3 70B
✓An open model for development, conversations, and your experiments.
Free routes
✳Reasoning
Qwen3
✓A flexible approach to reasoning, programming, and working with languages.
Free routes
✺Compact
Gemma 3
✓A lightweight open model for text, images, and quick prototypes.
Free routes
MGeneral purpose
Mistral Small
✓A practical model for conversations, code, and everyday work.
Free routes
iExample models. Availability and free limits depend on the provider.
FROM AN IDEA TO YOUR FIRST REQUESTStart creating.
Start creating.
Right now.
Install the router, add your provider key, and connect Claude Code. Everything after that is your next experiment.
- 01Install FreeLLMAPI
The app for Windows, macOS, or Linux.
- 02Connect a provider
Add an API key to your local gateway.
- 03Launch Claude Code
Choose a launch method in the terminal.
~/your-next-big-idea
# After setting up your local gateway
$
npx freellmapi launchFreeLLMAPI
✓Gateway ready to connect
✓Key passed only to the process
✓Your Claude Code, in a new flow
Launch example · not actual command outputNode.js ≥ 20.18 · configured gateway · Claude Code
# For macOS / Linux with Docker installed
$
curl -fsSL https://freellmapi.co/install.sh | bashFreeLLMAPI
→Installation script from the official project
→Gateway runs at localhost:3001
→Then add your provider keys
Is it really free?
The router itself is free and released under the MIT license. Free APIs have their own limits. The project also offers paid live catalog updates; the free catalog receives delayed updates.
Is this free Claude?
FreeLLMAPI connects the Claude Code interface to models from other providers. It does not offer free access to paid Claude Sonnet or Opus models.
What happens when every limit is reached?
The router tries the available routes. If every connected provider has reached its limit, you need to wait for the limits to reset or add another provider.
Can I choose an uncensored model?
Choose a model and provider that suit your experiments. Each provider defines its capabilities and request policies; there is no universal promise of uncensored access across all models.
Where are my API keys stored?
The gateway runs locally. Provider keys are stored with encryption. The launch command passes the gateway key to the Claude Code process without saving it in its configuration. Model requests go to your chosen provider.