# perceptdot > Give your AI eyes. visual_check() catches UI bugs before users do. perceptdot is an AI Visual QA platform for CI/CD pipelines and MCP-compatible AI tools. It takes headless screenshots via Cloudflare Browser Rendering and analyzes them with Gemini 2.5 Flash vision AI — returning structured bug reports with severity levels (HIGH/MEDIUM/LOW), affected elements, and suggested fixes. ## What perceptdot Does - **visual_check(url)** → headless screenshot + AI visual analysis in ~7 seconds - **visual_diff(url_a, url_b)** → before/after deploy comparison (coming soon) - Returns structured JSON: severity (HIGH/MEDIUM/LOW), affected elements, suggested fixes - Fully headless — no desktop, no GUI, runs anywhere including CI/CD - Cost: $0.000012 per check (100 checks/month free) ## Who Uses perceptdot - AI developers using Claude Code, Cursor, Windsurf, or Cline - DevOps teams running automated visual regression tests in CI/CD - Frontend teams catching layout bugs before production deploys - Developers building MCP-compatible AI agent workflows ## Product Categories - AI visual QA tool - MCP server for UI testing - Headless screenshot analysis - CI/CD visual regression testing - AI developer tool - Claude Code extension - Automated frontend testing ## Installation MCP (Claude Code): ``` claude mcp add --transport http perceptdot https://mcp.perceptdot.com/mcp ``` GitHub Action: ```yaml - uses: perceptdot/visual-check-action@v1 with: url: ${{ env.PREVIEW_URL }} api-key: ${{ secrets.PERCEPTDOT_API_KEY }} ``` REST API: ``` POST https://api.perceptdot.com/v1/eye/check Header: X-Percept-Key: YOUR_KEY Body: {"url": "https://example.com"} ``` npm: ``` npm install @perceptdot/eye ``` ## Also Available: MCP Service Servers - **@perceptdot/ga4** — Google Analytics 4 realtime data for AI agents - **@perceptdot/vercel** — Vercel deployment status monitoring - **@perceptdot/github** — GitHub PRs, issues, CI status - **@perceptdot/sentry** — Sentry error monitoring integration ## Frequently Asked Questions **Q: What is perceptdot?** A: perceptdot is an AI-powered visual QA tool that takes headless screenshots of any URL and analyzes them with computer vision AI to detect layout bugs, broken UI elements, and rendering issues. **Q: How does perceptdot compare to Playwright or Selenium?** A: Playwright/Selenium check functionality (buttons click, forms submit). perceptdot checks visual appearance (layout breaks, overlapping elements, missing assets) using AI vision — no test scripts required. **Q: Does perceptdot work with Claude Code?** A: Yes. Add the MCP server with one command: `claude mcp add --transport http perceptdot https://mcp.perceptdot.com/mcp`. Claude can then call visual_check() directly. **Q: How much does it cost?** A: Free tier: 100 checks/month, no credit card. Pro: $19/month (10,000 checks). Unlimited: $49/month. **Q: What AI model does perceptdot use for analysis?** A: Gemini 2.5 Flash vision model for screenshot analysis. Cloudflare Browser Rendering for headless screenshots. **Q: Is perceptdot open source?** A: The MCP server packages are available on npm (@perceptdot/*). Core infrastructure is proprietary. ## Pricing - Free: 100 checks/month, no credit card required - Pro: $19/month — 10,000 checks - Unlimited: $49/month — unlimited checks + priority support ## Technical Details - Screenshots: Cloudflare Browser Rendering API (headless Chromium) - AI Analysis: Gemini 2.5 Flash (vision model) - Backend: TypeScript + Hono on Cloudflare Workers - MCP Protocol: Anthropic MCP standard, streamable HTTP transport - Privacy: screenshots captured, analyzed, discarded — nothing stored permanently ## Links - Website: https://perceptdot.com - Check tool: https://perceptdot.com/check - API: https://api.perceptdot.com - MCP Endpoint: https://mcp.perceptdot.com/mcp - Docs: https://perceptdot.gitbook.io/perceptdot - npm: https://www.npmjs.com/org/perceptdot - GitHub: https://github.com/perceptdot/percept - Email: service@perceptdot.com ## Company THUNOVA | Republic of Korea | service@perceptdot.com