tashfeenahmed/freellmapi

What it solves

FreeLLMAPI provides a unified, OpenAI-compatible API endpoint that aggregates multiple free tiers from dozens of AI providers. It eliminates the need to manage multiple SDKs, API keys, and rate limits manually, allowing users to access a massive pool of free inference capacity (up to 7.4 billion tokens per month) across hundreds of model families.

How it works

The project acts as a local router that sits between your AI applications and various LLM providers. You add your provider keys to the router's encrypted SQLite database, and the router handles the request routing, automatic failover to the next available model when rate-limited, and per-key rate tracking to stay under free-tier caps. It also syncs a signed model catalog from a central feed to keep up with changes in the free-tier landscape.

Who it’s for

Developers and AI enthusiasts who want to maximize their use of free LLM tiers without the complexity of managing multiple accounts and providers, and those who want a single endpoint for their coding agents or AI clients.

Highlights

• Unified OpenAI-compatible surface: Supports chat, embeddings, images, video, and audio endpoints.
• Smart Routing & Failover: Six routing strategies with automatic retries and cooldowns for 429/5xx errors.
• Fusion (Multi-model synthesis): A virtual model that prompts multiple free models in parallel and synthesizes a final answer using a judge model.
• Broad Compatibility: Works with Claude Code, Cursor, Aider, and other OpenAI-compatible clients, and includes an MCP server for ChatGPT.
• Local-first & Secure: Provider keys are AES-256-GCM encrypted at rest and decrypted in-memory per request.
• Self-updating catalog: Automatically syncs new models and quota changes without requiring manual updates.
• Desktop App: Native menu-bar app for macOS and Windows for easy local management.