linuxhsj/openclaw-zero-token

What it solves

OpenClaw Zero Token removes the cost of using high-end LLMs by bypassing paid API keys. Instead of paying per request, it allows users to access models from providers like ChatGPT, Claude, Gemini, and DeepSeek for free by driving their official web interfaces through a unified gateway.

How it works

The system uses a browser-based authentication flow to capture session credentials. It launches Chrome in debug mode, allows the user to log in manually to the AI platforms, and then uses Playwright to intercept the necessary cookies and bearer tokens. These credentials are stored locally and used by a middleware layer to call the platforms' internal web APIs, mimicking a browser session. It also supports tool calling by injecting tool definitions into prompts, allowing models to interact with local tools like web search and file system access.

Who it’s for

It is designed for individuals who want to access multiple premium AI models without managing multiple paid subscriptions or API credits, as well as researchers experimenting with LLM capabilities across different providers.

Highlights

• Wide Provider Support: Works with DeepSeek, Qwen, Kimi, Claude, ChatGPT, Gemini, Grok, GLM, Xiaomi MiMo, and Manus.
• Unified Gateway: Provides a single HTTP API, Web UI, and CLI/TUI to interact with various models.
• AskOnce Feature: Allows a single query to be broadcast to multiple models simultaneously for side-by-side comparison.
• Local Tool Integration: Enables web-based models to execute local tools such as web_search, exec, and file read/write operations.