freellmapi: Route LLM Requests Through One API

Overview

FreeLLMAPI is a self-hosted gateway for sending requests to free LLM provider tiers and custom OpenAI-compatible endpoints through a single API. It aims to reduce the work of configuring separate provider integrations and handling their different limits and failures.

The router supports common LLM and media request types, offers a dashboard for managing provider keys and routing, and works with compatible clients and coding agents. It is intended for personal experimentation, not as a production inference service.

Key Features

• OpenAI-compatible API with additional Anthropic, Gemini, and optional Ollama-compatible interfaces.
• Routes across configured providers using model selection, quota tracking, key rotation, and automatic fallback on rate limits or server errors.
• Supports custom OpenAI-compatible endpoints, including local or remote services.
• Provides a dashboard for keys, routing chains, a request playground, and usage analytics.
• Encrypts provider keys in SQLite and decrypts them in memory for requests.
• Includes client setup helpers and an MCP server for compatible tools and agents.
• Can run locally or in Docker, with desktop installers also described in the README.

Use Cases

• Developers experimenting with free provider tiers who want to avoid wiring each provider into an application separately.
• Users of OpenAI-compatible SDKs or coding agents who want to switch among configured models through one base URL.
• Local-model users who want to route requests to their own OpenAI-compatible endpoint alongside hosted providers.
• Individuals prototyping chat, embeddings, image, audio, or video workflows while accepting free-tier limits and variability.

Getting Started

With Docker installed, run the installer:

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

Open http://localhost:3001, add provider keys, then use the unified API key and local /v1 endpoint. See the README for other installation options and client configuration.

Alternatives

• claude-code-proxy: Targets Anthropic-compatible clients and translates their requests to configured backends, rather than aggregating free provider accounts behind an OpenAI-compatible API.

Considerations

Free provider tiers can have changing quotas, variable latency, and limited availability; the README explicitly warns that this is for personal experimentation and learning, not production. You need provider keys to use upstream services, and their terms still apply when requests are routed through FreeLLMAPI. The optional live model catalog is a paid service, while the router itself is MIT-licensed.