How to Use FreeLLMAPI

In short: FreeLLMAPI is a free, open-source (MIT) router you run on your own computer. You add free API keys from providers such as Google, Groq, Cerebras, Mistral and OpenRouter, and it serves all of them from one OpenAI-compatible address, http://localhost:3001/v1, moving to the next model when one hits its limit. Its catalog lists 34 free providers, 635 free model endpoints and about 7.4 billion tokens a month of listed free-tier capacity. That figure is the sum of the providers' free tiers, not tokens from FreeLLMAPI: you only get what the keys you add allow.

What is FreeLLMAPI

The free version updates its model list from a monthly snapshot, so a new model reaches it 30 days after it is added to the live feed. The paid plan only makes that feed same-day; the router itself stays free.

Step 1: Install FreeLLMAPI

On Windows or macOS, download the desktop app from the repository's Releases page: the .exe installer for Windows, or the .dmg for macOS (arm64 for Apple Silicon, x64 for Intel). It runs from the tray and needs no password.

With Docker (Linux, macOS or WSL), run the one-line installer. It creates ~/freellmapi, generates an encryption key and starts the container:

curl -fsSL https://freellmapi.co/install.sh | bash

You can read the script at freellmapi.co/install.sh before running it. Running it again updates the container and keeps your keys.

Step 2: Add your free provider keys

1. Open the dashboard at http://localhost:3001 (on the desktop app: tray icon, then Open Dashboard).
2. Go to the Keys page and add a key for each provider you have. Free keys come from each provider's own site; aiblogly's free AI API keys guide has the steps for each one.
3. Reorder the Fallback Chain to choose which models are tried first.
4. Copy your unified API key from the top of the Keys page. It starts with freellmapi-.

Step 3: Use it from your code

Point any OpenAI client at the local address and the unified key. Use the model auto to let the router choose, or auto:fast, auto:smart or a specific model ID:

from openai import OpenAI

client = OpenAI(base_url="http://localhost:3001/v1", api_key="freellmapi-your-unified-key")
resp = client.chat.completions.create(
    model="auto",
    messages=[{"role": "user", "content": "Say hello in one sentence."}],
)
print(resp.choices[0].message.content)

Each response has an X-Routed-Via header naming the provider and model that answered.

Step 4: Connect a coding agent

FreeLLMAPI works with Claude Code, Codex CLI, Gemini CLI, Cursor, Cline, Roo Code, OpenCode, Aider, Kilo Code and others. Most agents set themselves up with one command, for example Claude Code:

npx freellmapi setup-claude --url http://localhost:3001 --api-key <unified-key>

The command backs up the agent's existing config before changing it. Any other OpenAI-compatible tool works with the base URL http://localhost:3001/v1 and the unified key.

Limits to know

• No frontier models and no uptime guarantee: it only reaches the free models of the keys you add.
• Daily caps: as top models hit their daily limits, answers come from smaller models later in the day; most limits reset at midnight UTC.
• Personal use: the project says it is for experimentation, not production, and each provider's own terms still apply to traffic sent through it.
• Single user: it is built to run locally for one person.

FreeLLMAPI vs OmniRoute

Both join free tiers behind one local OpenAI-compatible address. From each project's README: