Caveman Web

Ask supported AI chat sites to respond in terse Caveman-style prose. Local only.

Tired of AI models burying the answer in fluff? Caveman Web helps you ask for shorter, more direct replies by adding a small instruction to your prompt before it is sent.

When you press Enter or click send on a supported AI chat site, Caveman Web prepends an instruction such as:
“Use terse direct prose. No filler. Keep full technical accuracy.”

That is all it does. The model still decides how to respond.

Modes:
- Lite: gentle nudge toward concise answers.
- Full: classic Caveman mode: terse, accurate, no filler.
- Ultra: maximum compression, bullets only.
- Wenyan / 文言文: experimental classical-Chinese-flavoured terseness.

Caveman Web is honest about its limits:
- It is not a magic compressor.
- It does not guarantee token savings.
- It does not modify model output after the fact.
- It does not intercept network traffic.
- It does not reduce your bill automatically.

Privacy:
- No backend.
- No analytics.
- No telemetry.
- No third-party scripts.
- No chat storage.
- No network requests from the extension.

Your prompt text is read only long enough to add the selected Caveman instruction before sending. Settings and numeric usage counters stay locally in your browser.

Supported sites:
- ChatGPT: chatgpt.com, chat.openai.com
- Claude: claude.ai
- Gemini: gemini.google.com
- DeepSeek: chat.deepseek.com