Lead enrichment in Claude Code: people, emails and companies from 3,529 APIs

Four calls that turn a company into a contact

Lead enrichment is one motion: who works there, what is their email, is it real, and what is the company. Each step is a routed tool: treg.to runs the providers for that job, your own keys first, then the cheapest per hit, and the answer names the provider that served it. Claude Code can call them over MCP or run them from its shell:

# 1. who: people at a company, by the function you sell to
treg call treg.people.search --method POST --data '{"company_domain":"acme.com","title":"marketing"}'

# 2. their work email
treg call treg.people.email.find --method POST --data '{"full_name":"Jane Doe","domain":"acme.com"}'

# 3. is it deliverable: a found email is a best match, not a checked one
treg call treg.people.email.verify --method POST --data '{"email":"jane@acme.com"}'

# 4. the company: size, industry, location, from a bare domain
treg call treg.companies.enrich --method POST --data '{"domain":"acme.com"}'

Search for the buyer's function, not “decision makers”: in our 30 Sep run a decision-makers list for five accounts held 45 senior people and not one marketing title, while a title search for “marketing” found 20 people at all eight. Cap a whole waterfall with the X-Treg-Route-Max-Cost header, and run step 3 on every address before it goes into a sequence.

A verified lead list from one prompt

Hand Claude Code the ICP and let it chain the four calls, dropping companies before it pays for people:

Using treg, build a list of US B2B software companies with 51-200 staff that sell to marketing teams. Check each one against icp.md before any people lookup, find the head of marketing or growth, find and verify their work email, and write the rows to leads.csv with the provider that answered each step and what it cost.

A related, broader run on 23 Sep took 50 companies (48 with a usable domain) to 27 that passed the check and 20 verified deliverable contacts, for $2.33 metered including news lookups and opener scoring: $0.12 per deliverable lead. The 21 companies that failed the check never reached a paid lookup. The receipt is at find and verify a lead list, and the whole method, from ICP to rollout, is the GTM engineering playbook.

MCP, CLI or skill: which to use

• MCP for questions and small lookups. Claude Code sees catalog_search, catalog_get and call, reads the price before it spends, and the answer lands in the conversation.
• CLI for batches. A loop over 500 rows in Claude Code's shell writes each answer to a file, so the rows never pass through the context window, and the CLI prints what each call was charged.
• The skill is how Claude Code knows the flow. install.sh puts it in ~/.claude/skills/treg/; the same file is served at /skill.md.

Cost per correct work email, against the tools you would otherwise use

On the same 292 people (16 Sep 2026), cost per correct work email: treg.to $0.0056, Clay $0.0395, Freckle $0.0427, Deepline $0.0924. Exact match: 90.4%, 89.7%, 90.1% and 86.6%. Each column is an aggregator, not one provider. The method and every row are in the work email finding bench.