AEO Lookup

npx add-mcp 'https://aeolookup.com/mcp'

Connect an AI agent (MCP)

The server speaks MCP over streamable HTTP at https://aeolookup.com/mcp. It is stateless and answers in JSON, so any MCP client that supports remote servers can use it. Protocol versions: 2025-06-18, 2025-03-26, 2024-11-05.

Claude Code:

claude mcp add --transport http aeolookup https://aeolookup.com/mcp

Clients configured with JSON (Claude Desktop via a remote connector, Cursor, VS Code and others):

{
  "mcpServers": {
    "aeolookup": {
      "type": "http",
      "url": "https://aeolookup.com/mcp"
    }
  }
}

Tools on the server: check_ai_crawler_access, check_structured_data, validate_llms_txt, generate_llms_txt, score_answer_readiness, run_aeo_audit, calculate_ai_savings, calculate_ai_time_saved, estimate_ai_staff_impact, check_job_ai_risk, score_ai_readiness. Every tool is read-only and takes a url.

Call the JSON API

GET with query parameters, or POST a JSON body. GET https://aeolookup.com/api/aeo/ lists the tools. Responses are JSON with a summary, a score out of 100 where the tool scores, findings (each with a level of fail, warn or info) and the tool's own detail.

curl "https://aeolookup.com/api/aeo/ai-crawlers?url=example.co.uk"

curl -X POST https://aeolookup.com/api/aeo/audit \
  -H "content-type: application/json" \
  -d '{"url":"https://example.co.uk/"}'

Endpoint	MCP tool	What it does	Inputs
/api/aeo/ai-crawlers	check_ai_crawler_access	Checks whether the crawlers behind ChatGPT, Claude, Perplexity, Google AI Overviews, Copilot and 14 more may read a page: robots.txt rules (RFC 9309), robots meta and X-Robots-Tag directives, and whether the site's firewall refuses each crawler's user agent. Separates search and user crawlers (which decide whether you can be cited) from training crawlers (which do not).	url (required): The page or site to check, for example https://example.co.uk/ (https:// is added if missing). path: Optional path to test instead of the URL's own path, for example /blog/.
/api/aeo/schema	check_structured_data	Extracts every JSON-LD block on a page, lists the schema.org types it describes, flags missing required and recommended properties (Organization, LocalBusiness, Article, FAQPage, Product, Service, Person, BreadcrumbList, Event) and scores entity signals: a named organisation, sameAs links to other profiles, a logo and an @id.	url (required): The page or site to check, for example https://example.co.uk/ (https:// is added if missing).
/api/aeo/llms-txt	validate_llms_txt	Fetches /llms.txt (or the llms.txt address given), validates it against the llmstxt.org proposal (one H1, optional summary, H2 sections of [name](url): notes links), tests up to eight of its links and reports whether llms-full.txt exists. States the evidence on how little the file is read.	url (required): The page or site to check, for example https://example.co.uk/ (https:// is added if missing). check_links: Test up to eight of the file's links that are on its own site (default true).
/api/aeo/llms-txt-generate	generate_llms_txt	Drafts an llms.txt for a site: the name and summary from its home page, then its pages from the sitemap grouped by section, with titles and descriptions read from up to 24 pages. Returns the file as text, for a person to edit before publishing.	url (required): The page or site to check, for example https://example.co.uk/ (https:// is added if missing). max_links: Roughly how many links to include (default 40).
/api/aeo/answer-readiness	score_answer_readiness	Scores a page out of 100 on 14 published checks: indexable, title and description, one H1, question headings, direct first-sentence answers, structured data, dates, a named author, lists and tables, source links, specific figures, length, declared language and stock AI phrasing. Returns each check's points, what it found and the fix, ordered by points available.	url (required): The page or site to check, for example https://example.co.uk/ (https:// is added if missing).
/api/aeo/audit	run_aeo_audit	Runs all four checks on one address (AI crawler access, structured data, llms.txt, answer readiness) and returns each result plus an overall score: the mean of the crawler, schema and answer-readiness scores. llms.txt is reported but not scored, because the evidence says few systems read it.	url (required): The page or site to check, for example https://example.co.uk/ (https:// is added if missing).
/api/aeo/ai-savings	calculate_ai_savings	Estimates how much a UK business could save a year by handing admin to AI: a range in pounds from team size, weekly admin hours per person, hourly pay (default the ONS 2025 UK median) and sector exposure (Department for Education ranking), with two or three insights. Partial by design: the free AI audit gives the full breakdown.	industry: The business sector (default other). staff (required): Number of staff. admin_hours: Hours of admin each person does a week (default 10). hourly_pay: Average hourly pay in pounds (default 19.67, the UK median).
/api/aeo/ai-time-saved	calculate_ai_time_saved	Estimates the hours a week AI could save one person or a team, from the hours spent on seven everyday task types (email, scheduling, quotes and invoices, data entry, customer questions, documents, marketing), using a share per task informed by IPPR and McKinsey research. Returns a range and the task with the biggest saving.	email: Hours a week one person spends on email and messages (default 0). scheduling: Hours a week one person spends on scheduling, bookings and reminders (default 0). quotes: Hours a week one person spends on quotes, invoices and chasing payments (default 0). data: Hours a week one person spends on data entry, admin and updating records (default 0). customers: Hours a week one person spends on answering customer questions (phone, chat, email) (default 0). documents: Hours a week one person spends on reports, proposals and documents (default 0). marketing: Hours a week one person spends on social posts, newsletters and marketing copy (default 0). people: How many people spend their week like this (default 1).
/api/aeo/ai-staff-impact	estimate_ai_staff_impact	Estimates how much of a team's work AI could take on today, as full-time roles' worth of tasks, from headcount in seven role groups, ordered by the Department for Education's exposure ranking. Measures tasks, not people.	admin: How many people work as admin, reception and office support (default 0). customer: How many people work as customer service and phone sales (default 0). finance: How many people work as bookkeeping, payroll and finance admin (default 0). marketing: How many people work as marketing and content (default 0). professional: How many people work as professionals (accountants, solicitors, consultants, analysts) (default 0). managers: How many people work as managers and directors (default 0). handson: How many people work as hands-on staff (trades, care, kitchen, warehouse, drivers) (default 0).
/api/aeo/job-ai-risk	check_job_ai_risk	Rates a UK occupation's exposure to AI (high, medium or low, following the Department for Education's 2023 ranking), the share of its tasks AI can do today and the tasks it reaches first.	job (required): The occupation.
/api/aeo/ai-readiness	score_ai_readiness	Scores how ready a small business is to put AI tools to work, out of 100, from eight yes or no questions (digital records, one customer system, written processes, known repetitive tasks, current AI use, an owner, a clear website, time to bed in), and names the two biggest gaps.	digital: Most of your business records are in software, not on paper or in heads (15 points). onePlace: Customer details live in one system (a CRM, booking or accounts tool), not scattered across inboxes (15 points). processes: Your routine jobs (quoting, booking, onboarding, invoicing) follow written steps (15 points). repetitive: You can name the three most repetitive tasks in your week (10 points). using: Someone on the team already uses an AI assistant such as ChatGPT or Copilot every week (10 points). owner: One person is responsible for trying AI in the business (10 points). website: Your website answers customers' common questions clearly, so AI assistants can find and quote you (10 points). time: You could give a new tool two or three hours a week for a month to bed in (15 points).

OpenAPI 3.1 description: https://aeolookup.com/aeo-tools/openapi.json. A plain-text map of the site for language models: https://aeolookup.com/llms.txt.

Errors and fair use

• A bad or private address returns 400 with code E_INPUT or E_PRIVATE; a site that cannot be reached returns 502 with E_DNS, E_TIMEOUT or E_UNREACHABLE.
• The tools read public pages only. They refuse private, loopback and internal addresses, follow at most five redirects and stop after two megabytes or nine seconds.
• GET answers are cached at the edge for ten minutes per address. Each IP address may make 30 API requests and 60 MCP requests a minute; past that the server answers 429 until the minute is up.
• Our crawler identifies itself as AEOLookupBot/1.0 (+https://aeolookup.com/aeo-tools/), except for the firewall test in the crawler checker, which sends three answer crawlers' published user agents to see what your site returns to them.