Make your site agent-ready.
Get exact directions for your coding agent.
Get an AgentReady score, prioritized fixes, and prompts you can copy/paste into your coding agent to generate exact code changes.
Compatible with Cursor, GitHub Copilot, Claude Code, Codex, and any custom agent workflow.
No signup required for a public URL scan. Sign up for Enhanced Scan and GitHub repo checks.
Optional: connect GitHub for a full scan
Unlocks Code & Repo Quality checks (AGENTS.md, skills, CI/lint/tests, MCP hints, and basic secrets exposure signals).
You’ll authorize GitHub now, then create an account so we can save your connection securely.
What You Get:
Agent readiness score
A 0–100 score with a breakdown across protocol files, DOM & accessibility, content intelligence, browser agent compatibility, and security & safety (plus repo signals with GitHub).
Prioritized recommendations
Concrete fixes you can implement immediately, ordered by impact (high/medium/low) so you can ship fast.
Agent-ready prompts
Copy/paste a single “Agent Summary” prompt into Cursor/Claude/Codex (or any agent) to get exact, file-by-file fix instructions — or generate a prompt per recommendation.
What We Check:
Protocol Files
- •robots.txt + sitemap.xml
- •llms.txt + ai-agent.json
- •OpenAPI/Swagger discovery
DOM & Accessibility
- •Semantic landmarks
- •Labels + ARIA coverage
- •data-testid + agent hints
Content Intelligence
- •JSON-LD + schema types
- •Content structure signals
- •Form intelligence (types/autocomplete)
Browser Agent Compatibility
- •No-JS content signal
- •CTA/nav discoverability
- •Console errors + stability
Security & Safety
- •Security headers
- •CAPTCHA signals
- •Agent guardrails (data-agent-*)
Code & Repo Quality
- •AGENTS.md + skills
- •CI/lint/tests
- •Basic secret exposure signals
Enhanced Agent Tasks
- •Navigation intent
- •Safe form fill
- •Findability/search
How It Works:
Paste your URL
No code access needed
Optionally connect GitHub
Deeper scans for code & config
Review what’s missing
Score + prioritized recommendations
Ship with your agent
Copy/download prompts and implement
More than a score
Make your site reliably usable by agents (discoverable, navigable, safe to automate) — then ship fixes with your coding agent.
Agent onboarding
Make it easy for AI agents to discover rules, docs, and capabilities (robots/llms/manifests).
Reliable web automation
Reduce agent flakiness with better semantics, labels, and DOM hints (data-testid, data-agent-*).
Repo agent friendliness
Help coding agents contribute safely with AGENTS.md, skills, CI/lint/tests, and config signals.
Agent-driven fixes
Turn recommendations into prompts your coding agent can apply directly in your repo.
How we scan
Best-effort, real-world checks across your live site — plus deeper repo signals when you connect GitHub.
Protocol discovery
- robots.txt, sitemap.xml, llms.txt
- ai-agent.json + OpenAPI/Swagger discovery
- Bot-blocking signals (CAPTCHA/denials)
Playwright audit
- DOM semantics, labels, accessibility heuristics
- No-JS content, CTA/nav discoverability, stability
- Optional enhanced synthetic agent tasks (bonus)
GitHub Tier 2 scan
- AGENTS.md, skills, and instruction hygiene
- CI/lint/tests and agent tooling signals
- Basic secret exposure pattern scan
Tip: connect GitHub for Code & Repo Quality checks and more specific fix prompts.
FAQ
Quick answers to the common questions.
Do I need to connect GitHub?
No. You can scan any public URL. Connecting GitHub enables Tier 2 checks (AGENTS.md/skills, CI/lint/tests, MCP hints, and basic secret exposure signals) and improves prompt specificity.
Is this “real browsing” or a static scrape?
Scans use direct fetch + Playwright (Browserless when configured) to load JS-heavy sites and evaluate DOM, accessibility signals, and agent-automation heuristics.
Does Enhanced Scan affect my base score?
No. Enhanced Scan runs synthetic agent tasks and reports a separate bonus score so the base /100 remains stable.
Which coding agents does this work with?
Any: Cursor, GitHub Copilot, Claude Code, Codex, and custom agent workflows. We focus on giving clear instructions + prompts, not locking you into a tool.