Skip to content

Read more. Send nothing.

Summaries and plain-language explanations from a model running on your own device. No API key, and the model runs in your browser.

01 · The problem

Too many tabs, not enough hours.

Reports, threads and docs pile up faster than you can read them. Pasting them into a cloud chatbot is quick, but it means sending work documents somewhere you don't control.

02 · How it works

Three steps. All on your machine.

  1. Install a model once

    The Local AI page shows each model's size and its device fit before anything downloads. Chrome's built-in AI, in Chrome only, needs no download.

  2. Run it on the page

    Open the page and run Summarize This Page. Or run Explain This Page Simply for a plain-language version.

  3. Keep what matters

    Read the summary in the panel next to the page

The Summarize This Page recipe on the canvas: a start trigger, a step that reads the page text, an on-device summarisation step and the result displayThe Summarize This Page recipe on the canvas: a start trigger, a step that reads the page text, an on-device summarisation step and the result display
Summarize this page · the real recipe
What you'll need

Your device is enough.

  • Required
    The AWFlow extension
    Chrome, Edge, Brave or another Chromium browser; Firefox 140 or later also works.
  • Recommended
    WebGPU
    WebLLM chat models run on WebGPU. Where your browser or graphics card can't, the model shows Not supported here and you can pick another engine.
  • Not needed
    An API key
    On-device models are free to use and need no key.
Same recipe, other jobs

Same model, other jobs.

  • Explain

    Explain it simply

    Turn jargon into plain words on a page or a selection.

  • Rewrite

    Rewrite professionally

    Polish rough text in a professional tone with the Rewrite Professionally template.

  • Chat

    Chat with this page

    Ask follow-up questions about what you are reading.

04 · Questions

Fair questions.

Summarize recipe in the docs ↗
Do I need an API key?

Not for on-device models: they run on your hardware for free. You can still add a cloud provider with your own key for bigger jobs; your prompt then goes to that provider.

Is it as good as a cloud model?

Smaller on-device models handle summaries, explanations and rewrites well. For longer or harder tasks, a cloud model with your own key may give better answers.

What if my computer can't run it?

Each model card shows its size and device fit before you install. Chrome's built-in AI is an option in Chrome that needs no download.

Your reading list, finally readable.

Free to start, no key needed. Your first on-device summary takes minutes plus a one-time model download.