GTM Data Utilities: 17 Boring Endpoints That Stop Pipelines Breaking

Every enrichment script is mostly string handling. Split the comma-separated thing, strip the legal suffix, normalise the date, work out whether the URL redirects somewhere else. The expensive calls are five lines; the plumbing around them is two hundred.

These seventeen endpoints are that plumbing, exposed over HTTP. One page, because seventeen pages that each said "POST a string, get a cleaner string" would be seventeen pages of nothing.

One rule before the list: none of these are waterfall endpoints endpoints. Every one bills on any 2xx, including an empty or useless result. The miss-free behaviour covers email_finder, email_verifier and phone_finder and stops there. See waterfall email then verify for what that actually means.

They all share the same shape: POST https://api.richapi.ai/api/v1/{endpoint}, x-api-key header, JSON body.

/normalize_company: strips legal suffixes, punctuation and stray whitespace from a company name. normalize_case defaults to true. Reach for it when you're joining two lists on company name and one of them writes "ACME TECHNOLOGIES, INC." while the other writes "Acme Technologies".

curl -s https://api.richapi.ai/api/v1/normalize_company \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"company_name": "ACME TECHNOLOGIES, INC.", "normalize_case": true}'

/clean_domain: a messy URL in, a root domain out. Scheme, www., path and query gone. The dedupe key for a list of scraped URLs. It does not chain into company enrichment, and domain clean then company enrich explains that gap at length.

curl -s https://api.richapi.ai/api/v1/clean_domain \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"messy_url": "https://www.Example.COM/pricing?utm_source=x"}'

/normalize_list: splits a delimited string into a clean list, with remove_empty_values for the trailing-comma case. Useful when a form field or a CSV cell holds five values and your loop needs five rows.

curl -s https://api.richapi.ai/api/v1/normalize_list \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"input": "acme.com, beta.io,, gamma.co,", "remove_empty_values": true}'

/remove_whitespace: strips and collapses whitespace in input_text. Honestly: your language does this. It's here for agents and no-code tools with no string library.

curl -s https://api.richapi.ai/api/v1/remove_whitespace \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"input_text": "  Dana   Weiss  "}'

/count_occurrences: counts how many times separator appears in input, with case_insensitive and a limit. Sanity-checking a delimiter before you split on it, mostly.

curl -s https://api.richapi.ai/api/v1/count_occurrences \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"input": "a;b;c;d", "separator": ";", "case_insensitive": false}'

/encode_uri: RFC 3986 encoding of value. Same caveat as remove_whitespace: a one-liner in code, a genuine tool in a workflow builder.

curl -s https://api.richapi.ai/api/v1/encode_uri \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"value": "VP Engineering & Platform"}'

/format_datetime: parses a date string and reformats it, with original_timezone, new_timezone, locale and a strftime format. The one in this family that's genuinely annoying to write yourself, because timezone conversion plus localised month names is where date bugs live.

curl -s https://api.richapi.ai/api/v1/format_datetime \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"date": "2026-03-04T14:00:00Z", "original_timezone": "UTC",
       "new_timezone": "America/New_York", "format": "%B %d, %Y %H:%M", "locale": "en"}'

/distribute_leads: round-robins values_associated_with_labels across assignment_labels, with current_index so a later batch picks up where the last one stopped. This is lead routing without a routing product: pass the emails and the rep names, get the assignment.

curl -s https://api.richapi.ai/api/v1/distribute_leads \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"values_associated_with_labels": "alice@co.com,bob@co.com,carol@co.com",
       "assignment_labels": "Rep A,Rep B", "current_index": 0}'

Store current_index between runs. Without it every batch starts at Rep A and Rep A gets the lion's share.

/web_meta_tags: OpenGraph, Twitter Card and SEO tags from a URL: title, description, images, favicon. The cheapest way to find out what a company says it does before you spend enrichment credits on it.

curl -s https://api.richapi.ai/api/v1/web_meta_tags \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"url": "https://example.com", "cache": true, "cache_bust": false}'

/web_json_ld: Schema.org JSON-LD blocks from a page: products, articles, organisations. Structured data a site has already published about itself, which beats parsing its HTML.

curl -s https://api.richapi.ai/api/v1/web_json_ld \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"url": "https://example.com/product/widget", "cache": true}'

/find_redirect: follows link through every redirect to the final destination. Unmasking a shortened link, or catching an acquired company whose old domain now points somewhere else. That second case is a real signal.

curl -s https://api.richapi.ai/api/v1/find_redirect \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"link": "https://exmpl.co/x7f2"}'

/find_sitemap_urls: discovers a domain's sitemap and filters it by keywords, with exact_match. "Does this company have a /careers page, a /pricing page, an /integrations page" is a qualification question you can answer for one credit.

curl -s https://api.richapi.ai/api/v1/find_sitemap_urls \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"domain": "stripe.com", "keywords": "pricing,careers", "exact_match": false}'

/web_sitemap: every URL from a site's sitemap, via XML, robots.txt or crawl discovery. Per-result pricing: 1 credit. limit is required and you should treat it as a spending cap, not a convenience. A large site's sitemap is thousands of URLs, and at a credit each that is not a call you make casually.

curl -s https://api.richapi.ai/api/v1/web_sitemap \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"url": "https://example.com", "limit": 50, "cache": true}'

Use find_sitemap_urls when you have a question. Use web_sitemap when you need the whole thing and have priced it.

/directory_yellowpages: Yellow Pages business search by search_query and location, returning name, phone, address, website and categories. Per-result at 0.01 credits per result, with max_pages and page. The local-business route into a list, and the only endpoint on this page that generates rows rather than cleaning them.

curl -s https://api.richapi.ai/api/v1/directory_yellowpages \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"search_query": "HVAC contractor", "location": "Austin, TX", "page": 1, "max_pages": 2}'

/search_reference_data: 0 credits. Returns every valid filter label for lead search: seniority levels, industries, functions. No body required. Call it before you write a single filter; a guessed industry string returns an empty result set and no error.

curl -s https://api.richapi.ai/api/v1/search_reference_data \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" -d '{}'

/geo_id_search: 0.01 credits. Resolves a location name to a LinkedIn Geo ID for profile and lead search filters.

curl -s https://api.richapi.ai/api/v1/geo_id_search \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"search": "San Francisco"}'

Both feed people search to email waterfall, which is where they earn their keep.

It exists. It takes a first name and returns a predicted gender. 1 credit.

We're not going to sell it to you, because the accuracy limits are structural rather than fixable. Name-to-gender inference works by looking up how a name is usually assigned in some reference population, and that breaks in predictable ways: unisex names (Alex, Sam, Jordan, Andrea) have no right answer; the same name flips gender across languages; transliterated names from non-Latin scripts lose the information that would disambiguate them; and any name outside the reference population's distribution degrades badly. On top of all that, a person's gender is not a property of their first name, so even a perfectly accurate lookup of naming convention is still a guess about a person.

Which makes it a poor basis for personalisation specifically. The upside of getting it right is a pronoun nobody notices. The downside of getting it wrong is an email that tells the recipient you inferred something about them from a database and got it wrong. That's an asymmetric bet and it's the wrong side of it. Write copy that doesn't need the field.

curl -s https://api.richapi.ai/api/v1/predict_gender \
  -H "x-api-key: $RICHAPI_KEY" -H "content-type: application/json" \
  -d '{"name": "Dana"}'

If you're using it for aggregate analysis rather than per-person messaging (diversity reporting on a list, for instance), the error rate averages out somewhat and the inference is doing less harm. Per person, in an email, don't.

The honest warning about this whole page. remove_whitespace, encode_uri and count_occurrences are one-liners in every programming language. Calling them over HTTP inside a Python loop costs 0.5 credits a row plus a network round trip, to do something str.strip() does for free.

They belong in an agent's toolbox, or in a no-code builder with no expression language worth using. Inside a script, use your standard library and save the calls for format_datetime, find_sitemap_urls and the web extractors, which are genuinely doing work you'd rather not write.

The other way to burn credits here is an unbounded web_sitemap call, which is per-result at a credit each. Set limit.

Do any of these have a free-miss case? No. All seventeen bill on any 2xx. Only email_finder, email_verifier and phone_finder are waterfall.

Why is search_reference_data free? Because charging you to find out what the valid filter values are would be charging you for our documentation.

find_sitemap_urls or web_sitemap? Question versus inventory. The first is flat-rate and filtered; the second is per-result and complete.

Is directory_yellowpages US-only? It searches Yellow Pages, so treat coverage as whatever that directory covers in your target geography. Test before you plan a campaign around it.

Can I use these from MCP? Yes, all of them, same key and credit pool. Cursor and cursor enrich from csv show the agent-side pattern.

domain clean then company enrich for where clean_domain helps and where the chain stops, people search to email waterfall for the search endpoints these feed, n8n enrich and push hubspot for these calls inside a workflow, /platform for the endpoint catalog, and use cases for the rest of the plays. Tiers on pricing.

100 free credits, no card. That's two hundred of the 0.5-credit calls, which is enough to work out which of them you'd actually miss.