Pot

An open-source, cross-platform tool for selection translation and OCR on any OpenAI-compatible model.

What Pot is

Pot is an open-source, cross-platform tool for translating selected text and recognizing text in images (OCR). Among its translation engines is a built-in OpenAI service, so you can point it at TokenMix and translate with Claude, GPT or DeepSeek. It runs on Windows, macOS and Linux.

Before you start

• Install Pot from pot-app.com. It runs on Windows, macOS and Linux.
• Sign up at tokenmix.ai, open the Console, go to API Keys, and create a key that starts with sk-tm-.

Connect TokenMix

Pot's OpenAI engine has a Request Path field that is easy to fill wrong. Enter the full chat-completions URL, not just /v1.

1. Open Pot, go to Preferences, then Service Settings.
2. Under translation services, add the built-in OpenAI service.
3. Service Provider: keep it on OpenAI, not Azure.
4. Request Path: enter https://api.tokenmix.ai/v1/chat/completions, the full path. Do not enter just https://api.tokenmix.ai/v1, because Pot would turn it into /v1/v1/chat/completions and fail.
5. API Key: paste your sk-tm- key.
6. Model: enter a TokenMix model id, for example deepseek-v4-flash.
7. Save. Pot runs a quick test, and a pass means it is connected.

Check it works

Select any text and trigger Pot's translate. A translation in your target language means it is using TokenMix. You can see the spend in the TokenMix Console.

Common questions

Returns 404 or the test fails: the Request Path must end in /chat/completions. Filling just https://api.tokenmix.ai/v1 makes Pot append another /v1 and break. The bare domain https://api.tokenmix.ai also works, since Pot completes the path itself.

Which models to pick: for translation, fast models like DeepSeek V4 Flash, Gemini 3 Flash, Claude Haiku 4.5 and Qwen 3.6 Flash are a good fit.