A free ChatGPT proxy that learned generosity has rate limits

What it does This project is a thin middleware proxy—essentially glue code—that accepts ChatGPT API requests and forwards them to OpenAI using the maintainer’s own API key. It allowed developers to experiment with gpt-3.5-turbo without signing up for OpenAI or paying for usage. The source is open and designed for self-hosting if you supply your own key.

The interesting bit The public endpoint lived entirely on donated sponsorships, including 5000 CNY and 400 USD from contributors, which funded a shared API key for strangers until the maintainer archived the project in March 2023 citing an “unsustainable financial crossroad.” It is a remarkably honest post-mortem on why running a free API proxy is a credit-card-limit race against the internet.

Key highlights

• Drop-in proxy for the OpenAI ChatGPT API, accepting identical request formats and models.
• Rate limiting enforced to protect the shared API key from abuse.
• Privacy claims: the proxy itself does not log prompts, IP addresses, or user-agent strings (though OpenAI retains API data for 30 days).
• Can be self-hosted via Docker or Google Cloud Run using a personal API key.
• Served as infrastructure for downstream projects like the Free ChatGPT web app.

Caveats

• The public service is defunct. The repository was archived in March 2023 after the maintainer declared the free offering financially unsustainable.
• Running the public proxy relied on continuous donations; without sponsors, the free endpoint cannot exist.
• The project does not bypass OpenAI’s pricing—it merely shifts the cost to whoever owns the API key.

Verdict Worth a look if you need a reference implementation of a simple OpenAI API proxy or want to self-host a pass-through endpoint with rate limiting. Skip it if you are searching for a currently maintained free API or anything more sophisticated than request forwarding.