AI Settings — Bring Your Own API Key

Use your own LLM API key for the AI tools. Your key stays in this browser and is sent only to run a request — never stored on our server.

How it works

We provide a free shared model with low concurrency. When it's busy or exhausted, configuring your own key keeps the AI tools working. Your key is stored only in localStorage on this device and is proxied through our server (for CORS + safety) per request — we never persist it.

API Configuration

Provider / Format

Recommended for users in China — most providers are accessible without a proxy.

API Key

Model

Base URL (OpenAI-compatible endpoint)

Don't add /v1 or /chat/completions — we append them.

How to get an API key (tutorials)

Pick your provider below for step-by-step instructions. Providers marked ✅ work from mainland China without a network proxy.

AI Settings — Bring Your Own API Key

API Configuration

How to get an API key (tutorials)

OpenAI (ChatGPT)~$0.15 / 1M input tokens (gpt-4o-mini) — polishing one email costs well under $0.001.

Anthropic (Claude)~$0.25 / 1M input tokens (Haiku) — very low cost per polish.

DeepSeek (深度求索)~¥1 / 1M input tokens (deepseek-chat) — very low, and friendly to mainland China users.

Moonshot / Kimi (月之暗面)~¥12 / 1M tokens (8k) — long-context friendly.

Zhipu / GLM (智谱)glm-4-flash has a free tier — ideal for zero-cost trials.

Qwen / DashScope (通义千问)qwen-turbo is very low cost; Alibaba Cloud users can use it directly.

OpenRouter (multi-model aggregator)Hundreds of models, pay-per-use; some models have a free tier.