freellmapi: 34 Free LLMs Behind a Single OpenAI API

It unifies the free tiers of 34 different LLM providers into a single OpenAI-compatible endpoint. Instead of individual SDKs and complex rate-limit configurations, you simply point existing OpenAI client libraries to a local server. It exposes 474 model families and 635 free endpoints through a single /v1 API.

The router selects the optimal model for each request and automatically fails over to the next provider when one hits a rate limit. It tracks usage per key to ensure no free-tier limits are exceeded. The model catalog is automatically updated from a signed feed, keeping it current without git pull when new models are released or quotas change.

Major coding agents and CLIs such as Claude Code, Codex CLI, and Gemini CLI integrate directly with this API. Configuration is automatically generated via commands like npx freellmapi setup-claude, and existing settings are backed up. Zed and JetBrains AI connect via Ollama emulation.

While free tiers are individually limited, integrating them with this tool provides inference capacity of 7.4 billion tokens per month. It runs on Node 20+ and uses approximately 40MB of memory even on ARM SBCs like the Raspberry Pi. It is suitable for individual developers and small teams looking to reduce costs and experiment with various LLMs.