Full Self-Browsing
AI-powered browser automation through natural language. Tell it what to do, and watch it browse for you.
Built for the Modern Web
Everything you need for intelligent browser automation, powered by leading AI models.
Direct API Execution
Beyond clicking. FSB can call an app's own authenticated API through your logged-in session. Faster and more reliable than DOM automation, with DOM as the automatic fallback. Inspired by OpenTabs.
60 Browser Actions
Click, type, scroll, drag-and-drop, canvas drawing, CDP mouse control, multi-tab management, and more. 53 DOM actions plus 7 through the Chrome DevTools Protocol.
Reactive Triggers
Watch any element and get notified the instant a value changes, crosses a threshold, or matches a pattern.
Smart DOM Analysis
Incremental DOM diffing, multi-strategy selectors, and element identification for reliable automation.
Autopilot
Rebuilt on the Lattice SDK, long runs survive Chrome's MV3 shutdowns and finish. Up to 64 concurrent tabs, plus real file uploads from disk.
Visual Feedback
Glow borders, pulsing element highlights, and a floating progress overlay show exactly what FSB is doing.
Site Intelligence
130+ built-in site guides across 17 categories, from commerce to coding.
MCP Integration
Use FSB as an MCP server from Claude Code, Codex, Cursor, OpenClaw, or any MCP client. 66 tools, with browser automation straight from your IDE.
Remote Dashboard
Monitor and control FSB from any device. QR-code pairing, a live DOM preview powered by PhantomStream, now with media mirroring, and real-time task execution.
Memory & Learning
FSB remembers navigation patterns across sites. Semantic, episodic, and procedural memory for smarter automation over time.
Better Than All of Them. Without Vision.
Project Mariner, Claude Computer Use, OpenAI Operator, OpenClaw. They all rely on vision to understand the page. FSB reads the actual DOM. No vision needed. Faster. Cheaper. More accurate.
Project Mariner
VisionComputer Use
VisionOperator
VisionOpenClaw
VisionDOM-Based
Better than all of them. Pure structural intelligence. Zero vision. Zero guessing.
Works WITH All of Them
FSB does not just compete. It complements. Through its MCP server, Claude Code, Codex, Cursor, Windsurf, and any MCP-compatible agent can use FSB as their browser automation layer. The intelligence of any AI combined with the precision of FSB's DOM engine.
It clicks when it must. It calls APIs when it can.
FSB no longer has to drive a UI for everything. For supported apps it talks straight to the app's own authenticated API through your logged-in session, far faster and more reliable than DOM automation, and falls back to clicking automatically when no API exists.
How It Works
Three simple steps from instruction to completion.
Describe Your Task
Type a natural language instruction into the FSB chat panel. "Book a flight from SF to NYC" or "Find the cheapest laptop on Amazon."
FSB Analyzes & Acts
FSB reads the page, plans actions, and executes them. Clicking, typing, scrolling, and navigating as needed.
Task Complete
FSB verifies the result, reports back in the chat, and gives you a summary. Review the outcome and iterate if needed.
Choose Your AI Provider
Works with the leading AI platforms. Bring your own API key or connect LM Studio locally.
xAI Grok
Grok 4.1 Fast, Grok 4, Grok Code Fast 1
OpenAI
GPT-4o, GPT-4o Mini, GPT-4 Turbo
Anthropic
Claude Opus 4.6, Sonnet 4.6, Haiku 4.5
Google Gemini
2.5 Flash, 2.5 Pro, 2.0 Flash (FREE)
OpenRouter
200+ models through a single API key
LM Studio
Local models, no API key, live discovery from /v1/models
Start from FREE with Gemini 2.0 Flash, or run fully local with LM Studio and no API key.
Open Source. BSL 1.1 Licensed. Built for Everyone.
FSB is free and open source. Inspect the code, contribute features, or fork it for your own use.
View on GitHub