GLM Free API — Single File Bridge

GLM Free API is my most-starred repository, and the premise is almost rude in its simplicity: a complete API bridge to GLM in one Go file, with zero browser automation. No headless Chrome eating RAM, no puppeteer flakiness — just HTTP done properly.

The problem it solves

GLM's web chat is free but locked to a browser. Existing bridges wrapped that browser in automation, which meant memory blowups, captcha churn and crashes under load. I went the opposite way: reverse-engineered the web session flow, then implemented the protocol natively in Go. The result starts in milliseconds and scales to concurrent requests without a single browser process.

Features

• OpenAI-compatible /v1/chat/completions — drop-in for any existing client.
• Anthropic-compatible /v1/messages — so Claude-flavored tooling works too.
• Account pool — rotate multiple sessions, mark dead ones, stay under rate limits.
• Streaming — real SSE, not fake chunked responses.
• One file — go build, run, done. Read the whole implementation in one sitting.

One file is not a stunt. It is a statement that infrastructure should be readable — you can audit every line in ten minutes.

The same architecture spawned siblings: GhostBrain (Gemini), qwen-free-api and deepseek-free-api. Together with OmniRouter, they form a complete free-tier AI infrastructure that costs nothing to run.