For GTM engineers

Check every waterfall call instead of trusting a match rate.

68 data endpoints behind one key, over REST and MCP. The three waterfall endpoints, email_finder, phone_finder and email_verifier, return an execution_log on every call: which providers ran and why each one missed.

We don't publish match rates. Run a list you already know the answers for, and count.

Workflow

From a table to a scheduled job

Prototype the workflow in a spreadsheet or a Clay table. When it's settled, move it to a script or an n8n flow: the same endpoints, one key, a schedule.

Every endpoint's credit rate is in the OpenAPI spec, so you can price a job before you run it.

Signals

Data a cached database doesn't have

Some endpoints read live data at request time. These are the signals: things people and companies did this week, not a record last refreshed months ago.

Read live at request time:

  • Who commented on or reacted to a LinkedIn post
  • A person's recent posts and comments
  • LinkedIn job posts at a company
  • A company's ads in the LinkedIn and Meta ad libraries

Pricing

Pricing that doesn't punish automation

No seats: you pay for credits used, not for users. Search endpoints bill per result, so asking people_search for 100 people and getting 73 bills 73 profiles.

On the waterfall endpoints, a lookup that returns no result bills zero.

people_search · 0.1 credit per result

100Requested
73Returned
73Billed

= 7.3 credits

Run a fair test

Your cost per usable contact, in four steps.

  1. 01Take 200 contacts whose emails you already know are valid.
  2. 02Run them through email_finder without the emails.
  3. 03Count the results that match, and check the execution_logon the ones that didn't.
  4. 04Divide what you paid by the number of usable results.

FAQ

Frequently asked questions

A waterfall call that returns no result. It bills zero and says "billed": false. A call that returns a result bills at the published rate, so a wrong result still bills, and so does an "invalid" verdict from email_verifier. That's why step 3 of the test matters.
Yes: call it from any tool that can send an HTTP request with a header. It works well with TexAu, Clay, BitScale and others with HTTP or MCP support.
Yes. Connect the MCP server at mcp.richapi.ai/mcp. Credits come from the same balance.

Run the test on your own list.