Does an AI agent install your product correctly?
More and more people add tools to their sites by asking Lovable, v0, Cursor or Claude Code to do it. If the AI gets your install wrong, or builds its own version instead, you never hear about it. InstallGrade is my plan for a service that runs those installs for you and reports what broke.
What you can use today
None of these tests the install of a hosted product in AI builders yet, but each covers part of it.
| Tool | What it checks | Cost |
|---|---|---|
| agentic-usability (Nutrient) | How well AI coding agents use your SDK: it writes programming challenges from your SDK’s source, has agents (Claude Code, Codex, Gemini CLI or your own) solve them in isolated sandboxes using only your public docs, and has an AI judge score the results | Open source |
| AX benchmark (Tech Stackups) | Gives a coding agent (Claude Code) the task of integrating a service from scratch, and scores discoverability, sign-up, onboarding, API design and SDKs from 1 to 4 | Free to read |
| metaend Grade | Scans a URL and grades its “agent readiness” from A to F: discovery files, content negotiation, A2A, ERC-8004 and payments, with a fix for each failure | Free |
| agent-ready | Scores how ready a code repository is for AI coding agents (Claude Code, Cursor, Copilot, Codex) | Open source |
| agent-readiness | Benchmarks how agent-ready a code repository is for AI coding agents | Open source |
Checked October 31, 2026 from each tool’s own page.