Treg lets an AI agent find people by role, company, or location and return verified work emails, billed per result instead of a monthly seat. This launch wires that lookup to Jev, the decision model from TypeSafe, so the agent can keep only the contacts that match a job or target account before it pays for enrichment. It is meant for founders and operators doing outbound or recruiting inside Claude Code, Codex, and similar agents, as more of that list-building moves out of tools like Clay and Apollo. The hosted product at treg.to puts one token in front of providers such as Apollo, Hunter, Tomba, and People Data Labs, and meters each call at the provider's own rate with no markup. Work-email lookup is listed from $0.0089, and Treg says unsuccessful lookups are not billed. On treg.to, a published 100-lead test shows 92 correct work emails for $0.53, against about $3.35 and 84 correct emails for Clay. The same page reports that on LessieAI's People Search Bench, Claude Code with the plugin answered 78.2% of B2B prospecting tasks correctly, versus 43% for the same agent without it. Jason Zhou, who also founded Superdesign, is building Treg as a usage-priced catalog of data and tools for agents, with people search as the use case this launch is selling. The code is public at github.com/superdesigndev/treg, and new teams on the hosted service start with $1.00 of credit.