browsaBrowser Side-Panel AI

Quick start

Four steps: install the extension → open the side panel → connect your own model → ask your first question. About five minutes total.

1 · Install the extension

1

Chrome Web Store — easiest

Open the Chrome Web Store listing and click Add to Chrome. The store build updates itself, but store review can lag a few days behind the GitHub build.

2

Or: load from GitHub — newest features first

git clone the repo, or grab a zip from Releases and unpack it → open chrome://extensions (Edge: edge://extensions) → enable Developer modeLoad unpacked → select the browsa/ directory. Requires Chrome / Edge 114+ — any recently auto-updated browser qualifies.

Either way costs nothing, and browsa asks for nothing beyond what it needs.

2 · Open the side panel

On any page, press Ctrl+Shift+H (macOS: Command+Shift+H) or click the browsa toolbar icon. The panel opens on the right side of the page and follows whichever tab you switch to.

3 · Connect your own model

browsa ships with no model of its own — it brings the backend you configure into the panel. Click ⚙ Settings in the panel's top bar:

  • Plain chat models: any endpoint compatible with OpenAI, Anthropic, or Ollama. Fill in Base URL + API Key + Model ID, then Save.
  • Hermes Agent / OpenCode Agent / Agent Bridge: optional agent backends with tool execution, all built in — Agent Bridge connects your subscription-logged-in Codex / Claude Code — the model side runs on your subscription — plus pi with whatever model you configure it with. See Connecting backends.

Then hit Ping to verify — the first provider that pings successfully becomes active automatically.

No API key? Install Ollama locally (localhost:11434) for fully offline use, grab a key from any cloud provider's console — or use Agent Bridge to connect your subscription-logged-in Codex / Claude Code (subscription covers the model side) or pi (with the model you configure in it). Field-by-field details live on the Connecting backends page.

4 · Ask your first question

  1. Open an article (or a video, or a PDF), press Ctrl+Shift+H.
  2. Click 📎 next to the composer — the current page is attached, and a note above the input confirms what was read.
  3. Type your question and press Enter. The reply streams in token by token, with Markdown, formulas, and diagrams rendered right in the panel.

Don't feel like typing? Use the quick-action buttons above the composer — Summarize / Key Points / Explain / Outline — one click runs the task on the current page.

The browsa panel: provider switcher in the top bar, one-click quick actions, attachment and page meta pinned to the bottom
The panel: provider switcher in the top bar, quick actions one click away, attachment & page meta pinned to the bottom.
browsa's habit is "attach first, ask later." Once a page is attached it stays in context for every following turn — so click 📎 now and think about the question while the reply is streaming.

UI language

Settings lets you switch the interface to English / 中文 / Auto (follows the browser) — effective immediately, no reload. The language replies come back in can be pinned separately via the reply-language setting.