Clay AI agent benchmark: keeping subsidiaries off a target list

With Clay, Claude Code with Opus 5.5 built lists of 100 US-listed software companies with 91 to 94 meeting every rule, and Codex with GPT-6 Sol reached 82 to 88 for under 32 credits, because both checked each company's parent and headquarters before adding it. The method is a short checklist the agent runs on each candidate from Clay's company search.

By The Prompting Company · Snapshot November 5, 2026

The brief

We asked seven AI coding agents to use Clay to build a target list of 100 companies that meet three rules: they make software or provide IT or data-processing services, they are headquartered in the United States, and they are publicly traded on a US stock exchange. Each agent ran three times.

Every run that wrote a list built it from Clay's company search. The subsidiary question comes up because a well-known product brand, such as LinkedIn, can belong to a larger public company: a sales team targeting public companies needs the listed parent, not the brand.

Where does Clay's "Public Company" filter fit?

It is the right place to start. Clay's Find Companies search has a "Company types" filter that includes "Public Company", which narrows the search to the right kind of company. The top runs paired the search with a short check on each candidate before it went on the list, so the list holds listed parents headquartered where the brief says.

How the top agents built a clean list

Codex scored 88, 88 and 82 using only Clay data:

1. Search Clay's company data for software and IT companies in the US, paging through several thousand results.
2. Read each company's description, which often includes its exchange and ticker, to confirm it trades publicly in its own name.
3. Drop companies that are part of a larger group rather than listed themselves.
4. Fill missing HQ states with Clay's Enrich Company and Company Address functions, only for the rows that needed them.

Opus scored 93, 91 and 94. In its second run, the judge's review of the log shows it excluded acquired, delisted, foreign and non-software candidates before adding them.

Pass line 70. *Five domains came from the agent rather than Clay. †No list written.

A checklist for your agent

Ask the agent to apply these checks before writing each row:

• Is it listed in its own name? If the description or record points to a parent company, list the parent or skip the row.
• Where is the headquarters? Confirm the HQ country, not just a US office, and fill the state with Company Address if it is missing.
• Does it make software or provide IT services? Keep the companies whose main business fits your brief.
• Is it still trading? Keep companies that are listed today.

Accuracy came from these checks, and they cost very little: Codex spent 10 to 31 credits for 100 companies, under 1 credit per company, and scored 82 to 88. For the full task results see Can AI agents use Clay to build a target account list?

How we tested

The answer key. Whether a company is US-headquartered and listed on Nasdaq, NYSE, NYSE American or Cboe comes from SEC data as of November 3, 2026; whether it makes software or provides IT or data-processing services comes from its SEC industry code or a hand-reviewed classification.

The setup. Seven AI coding agents ran the task three times each: Claude Code with Opus 5.5 and Sonnet 5, Codex with GPT-6 Sol, Pi with GPT-5.6 Luna, Grok CLI with Grok 4.6, OpenCode with Kimi K3 and Factory with Kimi K3. Agents ran one at a time, each in a fresh sandbox with the same Clay key, the same 19 saved Clay routines and a 60-minute limit. A timeout counts as 0, and platform failures were rerun. Clay credits are measured as the change in the Clay workspace credit balance. A model judge read every log to confirm all values came from Clay. The task is at github.com/promptingcompany/clay-bench.

Try it yourself

The Clay Agent Plugin ships Clay's skills to your agent up front, including the habits that keep credit use down; setup is one prompt (about 15 minutes), and there is no additional cost to use it. Install the Clay Agent Plugin or give your agent a Clay API key and the Clay Public API docs. Ask it to build the list from Clay's company search with the Public Company type, then to confirm for each company that it is listed itself and headquartered where your brief says before adding it.

Common questions

How do I stop an AI agent from adding subsidiaries to a target account list in Clay? Tell it to confirm each company is listed in its own name before adding it; Opus and Codex, which ran that check, scored 82 to 94 of 100.

Is Clay's Public Company filter a good place to start? Yes. Start with Clay's "Public Company" company type, then have the agent confirm each company is listed in its own name and headquartered in the US.

What else should the agent check? That the headquarters is in the US, that the company makes software or provides IT services, and that it is still listed.