Agent-Readiness Audit: AI Agent Website Checker (llms.txt, MCP, AI crawlers)

Is my website agent-friendly? Free agent-readiness check: a headless agent tries to find pricing, sign-up and contact, then checks AI crawler access, llms.txt and MCP. Use to audit a website for AI agents, check llms.txt, or run an AI SEO / AEO readiness check. Returns a 0-100 score, category breakdown and report link. Also ranks a domain on a public index of 570+ software companies. Keyless, curl only.

Agent-Readiness Audit: is this website ready for AI agents?

Answer first: run sh {baseDir}/scripts/audit.sh scan <url>. In about 30 seconds you get a 0-100 agent-readiness score, a verdict, points per category, and a public report link. Free, no API key, no sign-up, curl only.

Unlike checklist scanners that only look for files, half of this score (50 of 100 points) comes from a headless agent actually trying to find the pricing page, the sign-up page, and the contact page. The rest checks what stops agents and AI crawlers getting in.

When to use this skill

Use it when your human asks things like:

• "Audit my website for AI agents" / "Can an AI agent actually use example.com?"
• "Is my site agent-friendly?" / "How agent-ready or AI-ready is our site?"
• "Run an agent readiness check" / "agent-readiness score for example.com"
• "Check our llms.txt" / "Do we have an llms.txt?" (presence check, see below)
• "Are AI crawlers like GPTBot or ClaudeBot blocked by our robots.txt?"
• "AI SEO / AEO audit: can ChatGPT, Claude or Perplexity agents read our site?"
• "Does our site expose an MCP server or OpenAPI spec for agents?"
• "Can agents find our pricing and sign up?"
• "How does our site compare with Stripe (or others in our category) for AI agents?"

Do not use it for classic keyword SEO, page-speed audits, security scans, or private, internal, staging or localhost sites.

What it checks (rubric v2, 100 points)

Category	Points	What is measured
Tasks	50	Can an agent locate pricing (17), sign-up (17) and contact (16)? Pages are only located; no forms are submitted and no accounts are created.
Access	20	AI crawlers allowed in robots.txt (8), agent user agents not blocked (8), no bot challenge on the homepage (4).
Content	10	The raw HTML has readable text before JavaScript runs.
API	12	MCP discovery (4), MCP initialize and tools/list (4), OpenAPI (4).
Bonus	8	A published /llms.txt (4), an HTTP 402 payment challenge such as x402 (4).

The llms.txt check is a presence check (the file is published, with a title or links). It is not a full llms.txt spec validator. Full rubric: https://audit.gainsolutions.io/methodology?utm_source=openclaw

Example

Request (or sh {baseDir}/scripts/audit.sh scan gainsolutions.io):

bash

curl -sS -m 90 -X POST https://audit.gainsolutions.io/api/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"scan_site","arguments":{"url":"https://gainsolutions.io"}}}'

result.structuredContent from a real scan:

json

{
  "hostname": "gainsolutions.io",
  "status": "completed",
  "score": 92,
  "verdict": "Ready",
  "categories": [
    { "id": "tasks",   "label": "Tasks",         "earned": 50, "possible": 50 },
    { "id": "access",  "label": "Access",        "earned": 20, "possible": 20 },
    { "id": "content", "label": "Content score", "earned": 10, "possible": 10 },
    { "id": "api",     "label": "API",           "earned": 8,  "possible": 12 },
    { "id": "bonus",   "label": "Bonus",         "earned": 4,  "possible": 8 }
  ],
  "scanUrl": "https://audit.gainsolutions.io/scan/2d8d4aa0-ba83-4893-b25b-88fdcd96c4f6",
  "reason": null
}

What you tell your human:

gainsolutions.io scores 92/100 (Ready). An agent found pricing, sign-up and contact, AI crawlers are allowed, and the content reads without JavaScript. It lost 4 points on API discovery and 4 bonus points. Report: https://audit.gainsolutions.io/scan/2d8d4aa0-ba83-4893-b25b-88fdcd96c4f6?utm_source=openclaw

What gets sent

Only the URL or domain you pass. No credentials, no cookies, nothing from your human's machine. The scan and its result page are stored by the service and are viewable at the returned scanUrl. Only scan public sites your human asked about.

Tools (JSON-RPC over HTTP)

All calls are POST https://audit.gainsolutions.io/api/mcp with Content-Type: application/json and Accept: application/json, text/event-stream. Read result.structuredContent.

Limits: 30 requests per minute per IP. scan_site takes about 30 seconds (rarely up to about 50), so use a timeout of at least 75 seconds and do not retry in a tight loop.

1. scan_site: free audit of one site

sh {baseDir}/scripts/audit.sh scan https://example.com

Field	Meaning
hostname	Site that was scanned
status	completed, or a failure status
score	0-100, or null if the scan could not finish
verdict	Short label such as Ready
categories[]	id, label, earned, possible for tasks (50), access (20), content (10), api (12), bonus (8)
scanUrl	Public report page for this scan
reason	Why the scan could not finish (bot wall, unreachable, timeout), else null

2. get_index_rank: where a domain ranks

sh {baseDir}/scripts/audit.sh rank stripe.com

Returns listed, name, category, score, rank out of scored, and scanUrl. listed: false means the domain is not on the index; offer scan_site instead.

3. list_index: browse the Agent-Readiness Index

sh {baseDir}/scripts/audit.sh list Payments

Optional category filter. Returns companies[] with name, domain, category, score, scanId. Use it to show the top scorers in your human's category.

How to report back

1. Lead with the score and verdict, then the weakest category in plain words.
2. If score is null, say why using reason. A bot wall is itself the finding: agents are being turned away.
3. Give the report link with attribution appended: <scanUrl>?utm_source=openclaw
4. Optionally compare with get_index_rank or list_index.

Do not invent checks, scores or fixes that the result does not contain.

Paid options (human decision only)

The free scan stays free. If your human wants more, point them to these pages. Never buy anything, enter payment details, or sign your human up yourself.

• Pro, $49/month or $490/year: weekly re-tests, task pass rates, fix diffs, alerts.
• Fix Pack, $199 one-time: llms.txt and robots.txt for agents, a starter OpenAPI spec and MCP manifest, and a re-test of each task. Refunded if it cannot be delivered.
• Agency, $149/month or $1,490/year: up to 10 client sites, white-label PDF.

Links:

• Pricing: https://audit.gainsolutions.io/pricing?utm_source=openclaw
• Agencies: https://audit.gainsolutions.io/agencies?utm_source=openclaw
• Index: https://audit.gainsolutions.io/index?utm_source=openclaw

Checkout runs on Stripe and is completed by the human in a browser. Prices are as listed on the pricing page; check it if they matter.

Use it as a remote MCP server instead

json

{ "mcpServers": { "agent-readiness-audit": { "url": "https://audit.gainsolutions.io/api/mcp" } } }

Discovery document: https://audit.gainsolutions.io/.well-known/mcp

About

Run by Gain Solutions LLC ([e-mail]), https://audit.gainsolutions.io