tl;air

About

tl;air reads the page you're on and tells you what matters — to you, not to a generic audience.

What you get:

- Personalized briefs. Paste your existing memory (ChatGPT export, Claude.ai notes, your bio), answer a 30-second chip picker, or upload Claude Code memory files. tl;air uses that context to pull out what's specifically relevant to your work, your projects, and your interests.
- Side-panel chat about the page. Ask follow-ups without leaving the tab. Click a citation and the page scrolls to the exact passage.
- Three layouts, strict length budgets. Classic TL;DR + bullets, or a three-section brief (takeaways / why this matters to you / what you can skip). Tight / Standard / Deep presets enforce real character caps so summaries don't balloon.
- Learns what you want. Thumbs up / down / refine per summary; per-site and global. Chat reflections auto-detect when a question reveals a real preference (and skip factual one-offs).
- Three auth paths, your choice:
  - Anthropic API key (BYOK)
  - OpenAI-compatible (OpenRouter, OpenAI, Ollama, LM Studio, any /v1/chat/completions endpoint)
  - Claude Pro/Max subscription via optional local helper
- Privacy-forward. No tl;air server, no telemetry, no account. Your keys and profile live in chrome.storage.local. Page content only leaves your browser when you summarize or chat — it goes to the provider you chose, nowhere else.
- Cache-first. Revisit a page, see the cached summary with "Saved X ago · Regenerate" instead of paying for it again. Prompt-caching on the stable system-prompt prefix keeps per-call cost low.
- Keyboard-first. Cmd+Shift+E toggles the panel, Cmd+Shift+K summarizes immediately. Configurable.

Best for: people who read a lot (research, engineering, investing, journalism) and want context-aware summaries instead of generic ones.

Privacy policy: https://wensenwu.com/privacy-policy

Note: tl;air is BYOK. You bring an API key (Anthropic, OpenRouter, OpenAI, or a local model). There is no subscription to tl;air.