PyBrowser

The browser that finishes things.

Browse normally. Ask Py anything. Give it research and multi-step jobs when you don't want to do them yourself.

Download See how it works

In active development — independent project

Py, the PyBrowser assistant, standing ready

Hi, I'm Py.

PyBrowser's new tab page, showing the PyBrowser wordmark, the Py mascot, and the prompt "What should Py do?"

Product

Three ways to work with Py

You can browse normally, ask me questions, or give me bigger jobs.

01

Browse

PyBrowser is a real desktop browser first — tabs, history, bookmarks, downloads, dark mode. Use it exactly like the browser you already know.

02

Ask Py

Ask about the page you're on, selected text, or your open tabs. Quick questions get quick, page-aware answers — no new window, no copy-pasting URLs into a chat box.

03

Missions & Research

Give Py a goal. It searches, reads, compares, keeps findings and sources organized, flags open questions, and checks with you before anything that matters — then hands back a structured result.

Ask Py

Available anywhere you're browsing.

Py sits in a panel next to the page, not in a separate app. Ask it things like:

  • “Summarize this page.”
  • “Explain this paragraph.”
  • “Compare my open tabs.”
  • “Research this topic deeper.”
  • “Which option is best?”

A quick question stays quick. A bigger one becomes a Mission.

The Py side panel answering a question about the page currently open, with Py's mascot shown in its reading state

Research

Research without the tab chaos.

Py can search across pages, review sources, and keep findings organized instead of leaving you with forty open tabs.

Py, the PyBrowser assistant

Searching…

Searching

Searching

Py opens the pages that look relevant to your goal, the same way you would — one tab, one source, at a time.

Reading

Every source gets reviewed, whether or not it turns into a finding. A page Py opened, read, and ruled out still counts as real work — you can see it was checked, not skipped.

  • ASICS Gel-Resolution 9 Review — useful
  • Nike Court Vapor 11 Review — skipped
  • Wilson Rush Pro 4.0 Review — useful

Thinking

Findings, constraints, and open questions build up as Py goes — nothing is thrown away, and nothing is invented to fill a gap.

Open Questions · 1

Reviewers disagree on how long the Nike Court Vapor 11's outsole actually lasts on hard courts.

Finished

The result is structured, not a wall of chat text — with the findings and sources it's built on still attached, so you can check Py's work.

A completed PyBrowser Mission: goal, constraints, a written result, two findings, one open question, and a list of sources marked useful or skipped

Missions

Give Py a goal.

A Mission is a persistent workspace for one goal — it keeps its own goal, constraints, findings, sources, and result together, and you can leave and come back to it.

Goal

Find the best hard-court tennis shoes under $120.

Constraints

  • under $120
  • hard-court durability
  • lateral support

Findings

  • ASICS Gel-Resolution 9 — $115, strong lateral support
  • Wilson Rush Pro 4.0 — $110, reinforced for hard courts

Sources

3 reviewed · 2 useful · 1 skipped

Result

The Wilson Rush Pro 4.0 is the best fit — $110, reinforced for hard courts, with strong lateral support.

Missions keep their goal, constraints, progress, findings, sources, and result together — so picking a task back up doesn't mean starting over.

Structured results

Not just another chat window.

What Py hands back matches the shape of what you asked for.

Comparison

Options weighed side by side against your constraints.

Ranked options

A short list, ordered, with the reasoning kept next to it.

Research summary

Findings and sources, organized under the question you asked.

Tab comparison

What's different — and what agrees — across the tabs you have open.

Safety & control

You stay in control.

PyBrowser asks before important actions — the kind that reach outside your own Mission board and actually change something. Py keeps working on everything else without interrupting you for it.

  • Submitting forms
  • Sending messages
  • Purchases
  • Deleting data
  • Sharing sensitive information
Py, the PyBrowser assistant, waiting for approval

I'll ask before anything important.

PyBrowser pausing before submitting a form and asking the user to approve or decline the action

Meet Py

The assistant inside PyBrowser.

Py is not the logo, and not decoration — Py's pose and expression are tied directly to what the browser is actually doing.

Py idle

Idle

Ready when you are.

Py reading

Reading

Going through a page or a source.

Py thinking

Thinking

Weighing what it's found so far.

Py working

Working

Acting on a step of the Mission.

Py waiting for approval

Waiting for approval

Paused before something that matters.

Py finished

Finished

Done, with a result ready to read.

Hover or select a state to see its motion. Motion is subtle by design — a breath, a lean, a small transition between states — and it stops entirely when your system asks for reduced motion.

Technology

Built with Python. Designed for AI-native browsing.

PyBrowser is a desktop browser built primarily in Python, with a Qt-based interface and Chromium-powered web rendering. Its AI systems are integrated directly into the browser, rather than living in a separate chat window — Py understands pages, navigates the web, manages Missions, and returns structured results from inside the browsing session itself.

Python

Core application logic, end to end.

Qt (PySide6)

The native desktop interface.

Qt WebEngine

Chromium-based web rendering — real pages, not a mock.

Multiple AI providers

Anthropic (Claude), Google Gemini, Groq, and OpenRouter.

Mission system

Persistent, structured multi-step tasks with their own store.

Safety layer

Action sensitivity is classified by the browser, not the model.

Engineering

Built from the ground up.

  • Missions persist their goal, constraints, findings, sources, and result across restarts.
  • Browser actions go through one structured, auditable API — not raw scripting.
  • History, bookmarks, and downloads are backed by a real local database.
  • Light and dark themes follow the desktop's own preference.
  • Covered by an automated test suite exercised on every change.

About

About PyBrowser

PyBrowser is an experimental AI-native desktop browser focused on making the web more actionable. Instead of only displaying pages, PyBrowser is designed to help you research, compare, organize, and complete work directly across the web.

PyBrowser is an independent software project, built in Python and currently in active development.

Preview Release · In active development

Download

Get PyBrowser

PyBrowser 0.1.0 is a preview release. Windows has been installed and used on a real machine; the macOS build passed its automated checks and is still pending a full manual pass.

Download for Windows Mac preview — test build

Both are unsigned preview builds — see the release notes for what your OS will show and how to open them. Prefer to build it yourself? View on GitHub.

Windows 10/11Unsigned preview build
macOSUnsigned · not yet notarized
LinuxRuns from source
Py, finished and ready to help

PyBrowser

So — what should I do?

Browse normally. Give Py the rest.

Download