Play 23
Browser Automation
AI-driven web navigation using vision + Playwright MCP.
Uses AI to navigate websites, fill forms, extract data, take screenshots, and execute multi-step web workflows — entirely driven by natural language instructions. Combines Playwright MCP Server for browser control (navigate, click, type, screenshot), GPT-4o Vision for understanding page content and making navigation decisions, and structured task planning for breaking complex web tasks into executable steps. Domain allowlist prevents arbitrary browsing.
Architecture Pattern
Browser automation: vision model + Playwright, task planning, domain-restricted
Azure Services
DevKit (.github Agentic OS)
- agent.md — root orchestrator with builder→reviewer→tuner handoffs
- 3 agents — Browser Builder (gpt-4o), Reviewer (gpt-4o-mini), Tuner (gpt-4o-mini)
- 3 skills — deploy (102 lines), evaluate (100 lines), tune (103 lines)
- 4 prompts — /deploy, /test, /review, /evaluate with agent routing
- .vscode/mcp.json — FrootAI MCP with OpenAI + target URL inputs + envFile
TuneKit (AI Config)
- config/openai.json — gpt-4o vision model, temp=0.1
- config/browser.json — domain allowlist, timeouts, viewport config
- config/guardrails.json — no credential entry, screenshot PII redaction
- evaluation/eval.py — Task completion >85%, Error rate <10%
Tuning Parameters
Machine evidence
FrootAI evidence lifecycle
This is an internal evidence maturity label, not third-party certification, accreditation, legal compliance, or a production guarantee. Missing or expired evidence demotes automatically; catalog claims cannot promote a play.
This play currently has design evidence only. A runnable scenario, endpoint evaluation, and build receipts are the next contiguous gates.
Repo Intelligence
v1A no-clone, revision-pinned map for agents and humans. Observed evidence is separated from inferred flow so the output stays useful without pretending to be a full call graph.