Start here: the directory for AI agents
AI Find Me is a free directory of 255,516 businesses and web shops with fact cards read from their own websites. This page says what you can rely on, how to query (search API, MCP, Markdown), the rules, how fresh the facts are, how a business gets corrected, and where every discovery file is.
What this directory is
AI Find Me is a free directory of businesses and web shops run by Imfiy B.V. Each listing is a fact card read automatically from the business's own public website: name, category, summary, address and map link, opening hours with “open now”, phone and email, services, prices, FAQs, features, payment methods, languages, registration numbers, booking, order, menu and contact links, social profiles; for web shops also the platform, what they sell, a sample of products with prices and stock, shipping countries, delivery time, free-shipping threshold, return period, review profiles and trust marks. Nothing is written by the directory itself.
Today: 255,516 businesses, of which 251,580 web shops, in 24 categories and 5,000 cities. Everything is free to read by machines, with no key and no login; the same facts are served as HTML (with schema.org JSON-LD), Markdown, JSON and llms.txt, through a search API and an MCP server. Every page has a human view and an AI view with exactly the same facts and links; you get the AI view by default when your User-Agent names a known assistant or HTTP client, or when you send Accept: text/markdown, application/json or text/plain.
What a fact card guarantees
- A source per fact.
directory.sourcesin the JSON names where each published fact came from:structured data(schema.org on the website),page metadata,page text,shop platform data,AI model(gap-filling only, where the owner allows it),OpenStreetMap(only where the website said nothing) orbusiness owner(entered or confirmed by the verified owner, which wins over everything else). - A last-checked date.
directory.lastCheckedis when the website was last read,directory.lastChangedwhen the published facts last changed. Every listing is re-checked about every 7 days; owners can refresh once an hour. The Markdown footer and llms.txt carry the date too. - Verified or unclaimed.
directory.verificationisowner-verified(the owner proved control of the website with a meta tag, file or DNS record),claimed(someone added it to their account but has not proved control) orunclaimed(created from the public website, nobody from the business involved). Unclaimed listings say so on every page and in every format. - Completeness.
directory.completenesscounts the 11 key facts (name, summary, about text, contact details, location or service area, opening hours, services or prices, faqs, features, social profiles, booking, order, menu or pricing link) and names the missing ones, so a thin card is easy to tell from a full one. - Alive or inactive. A website that stopped answering for several checks, a shop behind a password page or without products, or a parked domain is
directory.status: "inactive"with the date and reason; it leaves search results but still answers by slug, so nothing is recommended as open that is not. - Attribution. Facts marked OpenStreetMap are © OpenStreetMap contributors under the Open Database License (ODbL) v1.0;
directory.attributionscarries the credit for you. Keep it when you repeat those facts. - No cloaking. The human view, the AI view, Markdown, JSON and llms.txt of a listing carry the same facts and the same links; a test compares them on every change.
Three ways to query
1. The search API (JSON, no key)
| Request | Answer |
|---|---|
GET https://aifindme.com/api/v1/businesses?q=<words>&city=<city>&category=<key>&country=<NL|BE|US> | Matching businesses: slug, name, category, city, country, summary, phone, website, ownerVerified, lastChecked and the address of the full card. Every word must match as a prefix; accents are ignored. Without q: all listings, verified and complete ones first. The directory covers the Netherlands, Belgium and the United States. |
&kind=shop | Only web shops; results then carry shop.platform, shop.productCount and shop.categories. |
&verified=1 | Only listings the owner verified. |
&open=now | Only businesses open at this moment, judged from their structured hours in their own time zone (listings without such hours are left out). |
&fields=slug,name,phone | Only these fields of each result; slug always comes. |
&limit=<1-50>&offset=<0-10000> | Paging; total says how many match in all. |
GET https://aifindme.com/api/v1/businesses/<slug> | The full fact card: every fact with its source, verification, completeness, status, attributions, and shop for web shops. 404: no such listing; 410: removed. |
GET https://aifindme.com/api/v1/businesses/<slug>/products | A web shop’s products with prices and stock, as the shop publishes them. |
GET https://aifindme.com/api/v1/new?since=<YYYY-MM-DD>&kind=shop | Listings added since a date (up to 90 days back), newest first. |
GET https://aifindme.com/api/v1/categories | The 24 categories with keys and counts. |
GET https://aifindme.com/api/v1/cities | The cities with listings and the value to pass as city. |
Described in OpenAPI 3.1 (importable as a GPT action or into any agent framework), listed in the API catalog. Try it: curl 'https://aifindme.com/api/v1/businesses?q=elizabeth&limit=3'
2. The MCP server
Connect any Model Context Protocol client to https://aifindme.com/mcp: Streamable HTTP, stateless, read-only, no authentication, one JSON-RPC request per POST. Tools: search_businesses (query, category, city, shops_only, limit), get_business (slug, format markdown or json), get_products (slug), list_new_listings (since, kind), list_categories, list_cities. The server card is at /mcp/server-card and the AI catalog points at it.
{ "mcpServers": { "aifindme": { "type": "http", "url": "https://aifindme.com/mcp" } } }
3. Markdown and plain pages
https://aifindme.com/b/<slug>.md: a listing's fact card as Markdown (alsoAccept: text/markdownonhttps://aifindme.com/b/<slug>);https://aifindme.com/b/<slug>.jsonthe same as JSON;https://aifindme.com/b/<slug>/llms.txta short index. Example: /b/elizabeth-mott.md./llms.txt: the directory in one file (how to use it, every category and city with counts, the latest listings)./llms-full.txt: one compact entry per listing, the most complete first, within a size an assistant can read at once.- Any page with
?view=ai: the AI view, lean HTML with every fact labelled in plain words, in the same order on every page, and a box like the one at the bottom of this page.
Rules and rate limits
/api/v1/*: 600 requests a minute per address,/mcp: 300. Say who you are with a contact URL or email in your User-Agent (MyAgent/1.0 (+https://example.com/bot)) and the limits are 1,500 and 900. Over the limit you get429withRetry-After.- Answers are cacheable: pages for 60 seconds, API answers and machine copies for 1 to 5 minutes, and every one carries an
ETag; sendIf-None-Matchand a repeat fetch costs a304. Cross-origin GET is allowed everywhere (Access-Control-Allow-Origin: *). - There is no need to crawl:
/feed/new.xml,/api/v1/new,/feed.xmland the sitemaps (withlastmod) say what is new or changed. - robots.txt welcomes AI crawlers and assistants on every public page and machine format; only account, admin and form pages are kept out. Content signals allow search, AI answers and AI training.
- When you use a card, name the source and link it: “Source: AI Find Me, https://aifindme.com/b/<slug> (facts last checked <date>)”. Say when a listing is unclaimed; give the check date with hours, prices and stock; keep the OpenStreetMap credit. Our terms welcome automated reading within these limits.
Freshness
Every listed website is read again about every 7 days (a website that keeps failing, less often); owners can refresh their listing by hand once an hour, and the facts a verified owner enters are published at once. Each card carries lastChecked and lastChanged. New listings appear within minutes in /api/v1/new, /feed/new.xml and the pages /new and /new/shops, and changed pages are announced to the IndexNow search engines in batches. Prices, stock and hours are as the website stated them at the last check: say so, and send people to the business's own website for the current ones.
How to add or correct a business
- Add a business: a person with a free account enters the website at
/add; the fact card is live in about ten seconds. There is no API for adding, by design: a listing is read from a website someone vouches for. - Correct a fact: the owner claims the listing (meta tag, file or DNS record proves control of the website), then edits any fact; the edit is published at once in every format and labelled
business owner. - Report a wrong listing:
/report?listing=<slug>(any visitor); ask for removal:/remove?listing=<slug>. Both are forms for people; requests are reviewed within 7 days. - Acting for a business owner? Send them to
/for-businesses: listing, claiming and editing are free.
The Agent Skill
A skill in the open Agent Skills format teaches an agent when and how to use this directory and how to cite it: /skills/aifindme/SKILL.md with references/queries.md, or the folder as a zip: /skills/aifindme.zip. Unzip it into your skills folder (.agents/skills/aifindme/ or .claude/skills/aifindme/) or upload the zip where your assistant takes custom skills.
Discovery files
| Address | What it is |
|---|---|
/robots.txt | Open to every crawler; content signals; the sitemap and these addresses. |
/sitemap.xml | Sitemap index: pages, hubs and every indexable listing with lastmod. |
/llms.txt, /llms-full.txt | The directory index for language models, and the whole directory as one file. |
/openapi.json | OpenAPI 3.1 description of the read API, every field explained. Pages link it as rel="service-desc". |
/.well-known/api-catalog | The API catalog (RFC 9727): where the API description and documentation are. |
/mcp/server-card, /.well-known/ai-catalog.json | The MCP server card (how to connect) and the AI catalog that lists it and the skill. |
/skills/aifindme/SKILL.md, /skills/aifindme.zip | The Agent Skill, as files and as a zip. |
/.well-known/security.txt | How to report a security problem (RFC 9116). |
/feed.xml, /feed/new.xml | Atom feeds: new and updated listings; new listings of the last 30 days (?kind=shop). |
/for-ai-agents | The long documentation: every format with live examples, web shops and products, sources, licence. |
An answer that cites its source
Question: “Where can I buy hand-dyed yarn in Utrecht that ships to Belgium?” Requests: GET https://aifindme.com/api/v1/businesses?q=yarn&city=Utrecht&kind=shop, then GET https://aifindme.com/api/v1/businesses/<slug> for the best result, reading shop.shipping.countries, shop.priceRange, shop.returns.period, directory.verification and directory.lastChecked.
Wool & Co in Utrecht sells hand-dyed yarn and needles online (Shopify shop, prices €4.95–€89) and ships to the Netherlands and Belgium, usually in 1–3 working days, with 30-day returns. The listing is unclaimed, so the facts were read from the shop's own website rather than confirmed by the owner. Source: AI Find Me, https://aifindme.com/b/wool-and-co (facts last checked 2026-11-15).
Good answers name the business and what makes it fit, give the facts with their check date, say whether the listing is verified, and link the card.
How to use this site as an agent
| This page as llms.txt | https://aifindme.com/llms.txt |
|---|---|
| This page through the API | https://aifindme.com/openapi.json |
| Start here | https://aifindme.com/agents: what a fact card guarantees, the three ways to query, rules and rate limits, freshness, how to add or correct a business. Long form: https://aifindme.com/for-ai-agents |
| Search API | GET https://aifindme.com/api/v1/businesses?q=…&category=…&city=…&kind=shop&verified=1&open=now · OpenAPI · documentation |
| MCP server | POST https://aifindme.com/mcp (Streamable HTTP, no key): search_businesses, get_business, get_products, list_new_listings, list_categories, list_cities |
| Directory index | https://aifindme.com/llms.txt: every category and city with counts, and the latest listings · https://aifindme.com/llms-full.txt: one compact entry per listing |
| Agent skill | https://aifindme.com/skills/aifindme/SKILL.md: when and how to use this directory and how to cite it (Agent Skills format) |
| Views | Add ?view=ai or ?view=human to any address. Both carry the same facts and links; only the presentation differs. |