Skip to content

Introduction

Fantommind is an AI-powered browser automation desktop app for Windows.

It runs a local Python sidecar alongside a Tauri desktop shell, giving you a full browser automation platform without any cloud dependency — except for license validation at startup.

A profile is an isolated browser session with its own fingerprint, cookies, localStorage, and proxy settings. Profiles persist across runs so AI agents can maintain logged-in state.

A task is a reusable automation template: a prompt describing what the AI agent should do, the profile to use, the LLM provider, and optional configuration like max steps or timeout.

A run is one execution of a task. Runs record every step the agent took, screenshots at key moments, extracted data, and LLM token costs.

Schedules trigger tasks on a cron expression. They reload automatically when the app restarts — no separate scheduler process needed.

Webhooks post a JSON payload to a URL you configure whenever a run completes (succeeded or failed). Use them to notify Slack, trigger CI, or update your own database.

┌─────────────────────────────────────────────┐
│ Tauri Desktop Shell (Webview + React UI) │
│ localhost:1420 ──► sidecar :8765 │
└──────────────┬──────────────────────────────┘
│ HTTP / SSE
┌──────────────▼──────────────────────────────┐
│ Python Sidecar (FastAPI + APScheduler) │
│ SQLite WAL · AgentPool · LocalEventBus │
└──────────────┬──────────────────────────────┘
│ subprocess / CDP
┌──────────────▼──────────────────────────────┐
│ Browser Engine (Zendriver / Patchright) │
│ + browser-use AI agent layer │
└─────────────────────────────────────────────┘

The sidecar mounts all API routes. The Tauri shell loads the React dashboard as a local webview. There is no external server — everything runs on your machine.