Introduction to Recapt
Recapt is self-healing UX software. It detects friction in your product, diagnoses root causes, and fixes them automatically.
How It Works
Recapt creates a closed loop from problem detection to verified fix:
| Phase | What happens |
|---|---|
| Detect | Behavioral AI finds UX issues from real user sessions: rage clicks, dead clicks, form friction, confusing flows |
| Diagnose | Root cause analysis with session evidence and element-level friction data |
| Fix | AI proposes code changes, UX improvements, or copy fixes and writes them directly to your codebase |
| Verify | After deployment, recapt measures whether the fix actually improved user experience |
| Learn | Successful patterns are stored, making future fixes smarter |
This isn't analytics you have to act on. It's software that acts for you.
The Self-Improvement Workflow
Connect recapt to your AI coding assistant (Cursor, Claude Code, Windsurf, VS Code, Zed, or GitHub Copilot) and say:
Heal my site
The AI will:
- Find all UX issues across your site
- Investigate root causes using session data
- Write fixes directly to your codebase
- Track whether those fixes actually worked
No dashboards to check. No reports to read. Just better UX, continuously.
What Gets Fixed
Recapt detects all types of UX issues, not just bugs:
| Issue Type | Example | Example Fix |
|---|---|---|
| Rage clicks | Users hammering an unresponsive button | Add loading state, fix event handler |
| Dead clicks | Clicks on elements that look interactive | Add click handler, improve visual affordance |
| Form friction | Users abandoning forms mid-way | Improve labels, add validation, simplify |
| Confusing navigation | Users backtracking repeatedly | Add breadcrumbs, clarify CTAs |
| Console errors | JavaScript errors affecting real users | Fix the bug, add error boundary |
| Drop-offs | Users leaving at specific points | Reduce friction, improve content |
Behavioral Intelligence
Under the hood, recapt uses a custom behavioral model (not an LLM) trained specifically on how users behave when they're frustrated, confused, or struggling.
This model detects micro-signals that humans miss when watching session replays: subtle hesitations, tiny scroll adjustments, brief pauses that individually seem insignificant but collectively reveal user state.
Every session is automatically scored for:
- Frustration: rage clicks, dead clicks, scroll thrashing
- Confusion: hesitation, erratic movement, path wandering
- Confidence: direct navigation, purposeful interactions
Learn more: Behavioral Intelligence
Session Replay
When you need to see exactly what happened, recapt provides full session replay with:
- Complete DOM reconstruction
- Console errors and network requests
- User identification and metadata
- Privacy-preserving automatic obfuscation
Session replay is evidence for the AI to diagnose issues, and for you when you want to understand a specific user's experience.
Quick Start
1. Install the script (2 minutes)
<script
src="https://cdn.recapt.app/browser/recapt.js"
async
data-public-key="YOUR_API_KEY"
data-persist
></script>
2. Connect your AI assistant (5 minutes)
npx @recapt/cli init
3. Start healing
Heal my site
Full setup guide: Quick Start
Quick Links
- Quick Start: Get recording in 2 minutes
- MCP Setup: Connect your AI coding assistant
- Self-Improvement Workflow: The complete autonomous workflow
- Behavioral Intelligence: How the detection engine works
- Integrations: Get notified in Slack, Zendesk, Intercom