What it does

ChatAnywhere runs a reverse proxy for OpenAI-compatible APIs. You grab a free key, point your client at api.chatanywhere.tech (domestic China) or api.chatanywhere.org (global), and call GPT-5, DeepSeek, Claude, Gemini, Grok, and others using standard OpenAI SDKs. The repo itself is mostly documentation, status badges, and links to the actual service.

The interesting bit

The free tier is genuinely usable for light hacking: 200 requests/day on smaller models (GPT-4.1-mini, GPT-5-mini, GPT-3.5-turbo), plus a handful of calls on flagship models. For users inside China, the domestic endpoint avoids the friction of proxying through official APIs. The service also normalizes multiple vendors behind one OpenAI-shaped interface — swap the model string, keep the same code.

Key highlights

• Free tier: 200 req/day for mini models; 5/day for GPT-5/GPT-4o; 30/day for DeepSeek-R1/V3
• Paid tier claims lower latency, no rate limits, and per-token billing “below official prices”
• Domestic China acceleration without VPN required
• Supports Anthropic native protocol for Claude models alongside OpenAI-style calls
• Status page and uptime badges for each model family

Caveats

• Free keys are personal/non-commercial only; “commercial or mass-facing use” is explicitly at your own risk
• Rate limits are enforced per-IP-and-key combined, so sharing keys across machines or IPs still caps at 200/day
• The project warns that free GPT-5 series responses have “weaker reasoning ability” than paid tiers
• Keys may be banned for abuse; recovery is via a QQ group, not automated

Verdict

Worth a bookmark if you need quick, no-card API access for personal scripts or prototyping inside China. Skip it if you need reliability guarantees, commercial licensing, or don’t want to route your prompts through a third-party relay with opaque infrastructure.