Turning free chatbot frontends into an OpenAI-compatible API

What it does gpt4free-ts is a TypeScript port of the gpt4free project. It acts as a middleman between your application and a long list of third-party websites that offer free LLM access, repackaging their responses into a familiar chat-completion format. You pick a backend “site” like you or phind, and it forwards your prompt through that site’s unofficial frontend.

The interesting bit The value is entirely in the glue: it normalizes dozens of disparate, undocumented web frontends into a single v1/chat/completions interface. The README openly admits this is a cat-and-mouse game—many reverse-engineered targets are already offline, and the author invites sites to request removal.

Key highlights

• Supports a wide range of models depending on the target site, including gpt-4, gpt-3.5-turbo, Claude variants, Llama 2, and Mixtral.
• Offers both standard request/response and Server-Sent Events streaming endpoints.
• Provides a publicly hosted API key and base URL, though the project is clearly intended for self-hosting via Docker.
• Includes a /supports endpoint to query which target sites and models are currently mapped.

Caveats

• The target site list was last updated in September 2023, and the author notes that “most of the sites here are no longer available.”
• The project is strictly framed as “educational purposes only” and disclaims any association with the underlying API providers; using it likely violates the terms of service of the scraped sites.
• Stability is entirely dependent on third-party frontends that can change, break, or block access at any time.

Verdict Worth a look if you need a quick, local proxy for experimenting with free LLM frontends, but avoid it for anything production-critical or legally sensitive. If you want reliable GPT-4 access, this is not your infrastructure.