Surfing Dog Search

Checks what an AI agent can do with a business's website: message, book, order, sign up, pay, change or cancel. Waits up to 25 seconds; if the read isn't finished, returns what it found so far, and you call it again for the score. A result read in the last hour is answered as it is; after a business changes its site, set fresh to true to read it again (allowed a few minutes after the last read). Use it when a person asks how agent-ready a business is, or wants to improve their own. To find businesses, use search_businesses instead.

One business's full listing, by its domain: its address (a town and country for one this network found), the week's opening hours and closures, its doors and what each takes, its services, its standing in words, a member's track record (promises kept, broken and still open, and when, once enough have closed; self_reported when it rests on the business's own word, with no score and no count of customers), and how many of its promises closed with each outcome; for one that is not a member, every fact with its source and date. This server takes no bookings or orders; the listing names the business's own doors. Its agentic score, if checked, is at https://surfingdog.ai/b/.

The attributes a search may ask for (delivery, walk_ins, wifi…), each with its labels, and the payments accepts takes. Pass keys as attributes to search_businesses, and payments as accepts.

The categories list: every slug with its English and Portuguese labels and its kind (local_and_retail, or software_and_ai for software, API and AI companies, kept apart). Pass a slug as category to search_businesses.

For a business's own AI: put the business in this directory with its doors (MCP, A2A, OpenAPI, an API, UCP, ACP, a platform's door) and its facts: name, what it is (a category of list_categories or a place category, and one line), where (a storefront, a service area, or online with the countries it ships to), its languages, and agree. One call when its proof is already in place; otherwise the answer gives a challenge to put on its domain (a well-known file, its inbox manifest, a DNS record), a key's signature, or a code sent to an address at its domain, and the next call proves it. The answer names every part not taken and why, its readiness level, and what the next level takes. Doors count once this network's crawler has checked them; a door on another host counts only when the site leads to it or its own document names the business's domain. Mail, phone, messaging, forms and web pages are never doors. Every kind of business is listed except adult services, gambling, weapons, tobacco, cannabis and vaping, payday loans and bail bonds. agree accepts the listing terms (/bot#terms). Being listed is free, and nothing about the order can be bought.

Find businesses in the directory by what they do (query, category: a group or a place category), where (near, country), the language they speak, whether they are open now, their attributes (list_attributes), price band, what their doors take and what they accept (accepts), whether a door answers asks or quotes (requestable), the type of door, their readiness level, whether they have an inbox, and whether they are members, registered or found. Results come in the directory's published order, which nobody can pay to change; every filter only leaves businesses out. This server takes no bookings or orders; each result lists the business's own doors. A business this network found on its own website is marked as found, with the date it was checked. Software, API and AI companies (category software-ai) are kept apart: they come only when category names that group or one of its categories, query names their kind, or every word of query is in one's name. Pages of up to 20; pass next_cursor as cursor for more.

For a business's own AI: correct its listing (set any field register_business takes), add or remove doors, turn the listing off or on, or hide its agentic score result (score_page), with its claim_token or a proof as register_business asks. opt_out stops this network listing the site (list), reading it (crawl), or both (all). With no proof it removes only a listing nobody claimed; a claimed or registered listing is taken out only with its claim_token or a proof as strong as its claim.