LazeeBrowse.ai drives the browser you already have open — navigating, clicking, typing, running JavaScript, and capturing network activity — while producing reports, PDFs, DOCX/XLSX files, and more.
Everything the agent can do out of the box.
Navigate, click, type, scroll, submit forms, and press keys through a live Electron webview — the agent interacts with the browser exactly like you would.
Execute arbitrary JavaScript in the live page context to scrape data, mutate the DOM, or drive embedded app logic directly.
Inspect the payloads a site loads — URLs, methods, status codes, and request bodies — via Electron's webRequest hooks.
When a task is complex, the agent explores multiple strategies, scores each one, backtracks, and replans until it converges on a solution.
Produce reports as PDF (HTML/CSS via headless Chromium), DOCX, XLSX, or text/markdown — and read back PDF, DOCX, XLSX, and PPTX files.
Recent messages are kept raw while older context is rolled into a summary persisted in SQLite, so long sessions stay coherent.
Bring your own API key for OpenAI, Anthropic Claude, Google Gemini, DeepSeek, or OpenRouter — switch models from a dropdown any time.
Clean frameless Electron window with light/dark themes, a file sidebar, and a save-to-PDF overlay for capturing results.
Two lightweight pieces talking over a WebSocket.
Hosts the live browser webview, captures network traffic, executes JS, and manages the UI. It also spawns the backend at launch.
A WebSocket server that runs the agent loop: planning, tool selection, Tree-of-Thought deep dives, and chat/memory management.
Each session connects to ws://localhost:8000/ws/{session_id}. Browser commands flow as browser_command messages and return as browser_result.
The agent saves generated files, reports, and summaries to your chosen project folder — reviewable and editable afterward.
LazeeBrowse.ai in action.
Prerequisites: Python 3.13+, Node.js 18+, and an LLM API key from any supported provider.
# One-command setup and launch
./run.sh
# — or manually —
# Backend
cd fastapi-server
python3.13 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python main.py
# Frontend (separate terminal)
cd electron
npm install
npm startPrefer containers? docker compose up --build desktop runs the full app over noVNC at http://localhost:6080/vnc.html.
Open source throughout.