Agent Ready — Cursor Plugin

Scan any URL for AI agent readability directly inside Cursor — against the Vercel Agent Readability Spec, the llmstxt.org standard, and seven agent-protocol manifests (MCP server cards, A2A agent cards, agents.json, agent-permissions.json, UCP, x402, NLWeb). 60 checks across four spec families with per-check howToFix guidance.

Backed by agent-ready.dev — the same engine behind the REST API, the MCP server, and the browser extension.

What's in the bundle

Component	What it gives you
MCP server (mcp.json)	The agent-ready server with three native tools: scan_site, get_scan, ask. Plus three workflow prompts: scan, interpret_scan, remediation_plan.
1 skill (skills/agent-ready/)	"When and how to scan" guidance: tool selection, the running placeholder, summarisation rules, the four check families (S/P/L/C), and what not to use Agent Ready for.
1 rule (rules/agent-ready.mdc)	On-demand tool-selection guidance: which tool for which question, plus the rules of engagement (URL verbatim, no fabricated scan ids, don't dump raw JSON).

Install

Install from the Cursor Marketplace (search "Agent Ready"), or test locally with the included sync script:

./scripts/sync-local.sh

That copies the plugin into ~/.cursor/plugins/local/agent-ready/. Then Cmd+Shift+P → Developer: Reload Window in Cursor.

Configure your API key

scan_site and get_scan need an Agent Ready Pro API key. The ask tool is public — no key.

1. Issue a Pro key at https://agent-ready.dev/dashboard/api-keys.
2. Set AGENT_READY_API_KEY in your shell, or put the literal key in ~/.cursor/plugins/local/agent-ready/mcp.json (never the repo copy).

Dock-launched Cursor doesn't inherit shell env. If ${AGENT_READY_API_KEY} resolves to empty when Cursor is launched from the Dock, either launch from a terminal or paste the literal key into the installed mcp.json.

{
  "mcpServers": {
    "agent-ready": {
      "command": "npx",
      "args": ["-y", "agent-ready-mcp@latest"],
      "env": { "AGENT_READY_API_KEY": "${AGENT_READY_API_KEY}" }
    }
  }
}

Try it

Once installed, ask Cursor's agent things like:

• Scan https://example.com for AI agent readability and summarise the top failures.
• What does check P11 measure, and how do I fix it?
• Fetch scan abc1234567 and produce a remediation plan focused on agent protocols.
• Is my /.well-known/mcp.json valid?
• Search Agent Ready's methodology for "markdown mirror". (ask, no key needed)

Rate limits

• Pro: 10 scans/minute, 200/day. Read endpoints (get_scan) 120/minute.
• Team: 60 scans/minute, 2000/day.
• Public ask: rate-limited per IP, no key required.

Links

• Website: https://agent-ready.dev
• Methodology: https://agent-ready.dev/methodology
• Check registry: https://agent-ready.dev/checks
• Agent guide: https://agent-ready.dev/AGENTS.md
• OpenAPI spec: https://agent-ready.dev/api/v1/openapi.json
• MCP server source: https://github.com/mlava/agent-ready-mcp
• Browser extension: https://agent-ready.dev (footer)

License

MIT — see LICENSE.