The GTM MCP Directory
Directory / Learn / How to actually do it
How to actually do it

How do I build a prospect list with an AI agent?

The short answer

The chain is: define the account criteria, search a company database, enrich each company, find the right people inside them, then score and prioritise. Each of those is a separate job in this directory with its own coverage, and the account list step is the thinnest link.

The chain

StepThe jobTools taggedOfficial MCPFree tier
Build the target account listbuild a target account list740
Search companies by firmographicssearch companies by firmographics15123
Enrich each company from its domainenrich a company from a domain342612
Search people by criteriasearch people by criteria24155
Score and prioritisescore and prioritize leads24153

Counted 2026-08-25. Entry counts rather than product counts.

Where it is thin

build a target account list is tagged on only 7 entries, of which 4 have an official server. In practice this step is usually done by combining a firmographic search with a signal rather than by a tool sold as list building, which is a reasonable thing for an agent to orchestrate itself.

The tools with the deepest coverage in the chain

5 more are on the linked page. The cut is the display limit, not a ranking.

Add a signal, or you have built a directory rather than a list

A list of companies matching a size and an industry is not a prospect list, it is a phone book. The difference is a reason to reach out now. 12 entries are tagged detect a funding or news event, 5 are tagged scrape job postings, 14 are tagged detect a company's tech stack and 10 are tagged track job changes. Those are the cheapest sources of a real reason.

Then stop

Building the list and sending to it are different decisions. Keep the send behind a human approval until you have read what the chain produced at least twice, because an agent will build a list of four thousand as easily as forty and will not feel embarrassed about either.

Sources

Every number on this page is generated from directory.json at build time and carries the date it was baked: 2026-08-25. Nothing is typed by hand, nothing is rounded, and nothing is estimated. The underlying data is published in full. Where the honest answer is a zero, the zero is printed.

Related questions

In the directory