Jonathon's AI Wiki

Clay + Claude Code — Natural-Language Lead Generation (Nate Herk)

4 min read

Source: raw/Claude_Code_+_Clay_Makes_Lead_Generation_Actually_Fun.md — Nate Herk walkthrough (youtube.com/watch?v=zyvdl__Ywfk) of driving Clay’s B2B data engine from inside Claude Code.

Clay ships an official Claude Code plugin + MCP server, which turns lead sourcing, waterfall enrichment, and cold-email copywriting into a single natural-language /goal run — no Clay UI table-building required. Herk’s demo produces a campaign-ready CSV of 50 enriched HVAC/home-services decision-makers with personalized subject lines and email bodies, then hands off to Clay’s own sending infrastructure. All run figures are the creator’s single demo, not benchmarks.

Key Takeaways

  • Install is marketplace-native: /plugin → add marketplace → Clay’s marketplace link → install → reload, done from the Claude Code terminal (not the desktop/VS Code app). Then “help me authenticate” loads Clay’s onboarding skill and connects the Clay MCP server.
  • MCP tool surface (credit-metered): find email, enrich person, enrich company, lookup phone, read/edit Clay tables, validate workflow — each call consumes Clay credits.
  • The one-prompt workflow: a /goal spec — 50 enriched HVAC/home-service leads, decision-makers, email + pain points + subject line + body, CTA = “yes to a 90-second Loom,” no blank columns, deliver CSV — makes Claude fan out six city sub-agents (Houston, San Antonio, Atlanta, Charlotte, Tampa, Las Vegas), enrich, dedupe, run verification/debate passes, and emit a 50-row CSV.
  • Waterfall enrichment is the reason to pay Clay: chaining data vendors lifts email hit-rate from ~30% (single vendor) to ~80-90% (creator-cited Clay figure).
  • Cost/time of the verified run: **172 Clay credits ≈ 24/100); the verification-heavy run took ~1 hour vs ~5 minutes for a plain 50-lead pull.
  • Context prerequisite: the copy is only as good as the project context — Herk loads business-profile, case-studies, FAQ, proof, offer, and website-copy files into the Claude Code project (plus a prior cold-email podcast transcript as the copy playbook) before the run.
  • Downstream (still in Clay’s UI): import the CSV → build the campaign with / row variables for subject/body → buy and auto-warm sending domains inside Clay (5 accounts per domain, 30 sends/day). Stated limitation: Clay’s MCP cannot yet manage the campaign/follow-up send step.

Try It

  1. Install the plugin from the Claude Code terminal: /plugin → add marketplace → Clay’s marketplace URL → install → reload → “help me authenticate.”
  2. Stage your context files first — business profile, offer, proof/case studies, tone examples. The enrichment is commodity; the non-generic copy comes from this folder.
  3. Copy the /goal shape: lead count + ICP + geography + required columns (email, pain points, subject, body) + CTA + “no blank columns” + output format. Let Claude fan out sub-agents and verify.
  4. Budget with the demo’s meter: ~$24/100 verified leads in Clay credits, plus roughly an hour of agent runtime per 50 verified leads — then A/B the copy in Clay’s campaign builder.
  5. Keep the send step in Clay (domain buying, warm-up, sequencing) — the MCP doesn’t reach it yet.

Open Questions

  • Clay credit pricing tiers and what plan the 172-credit run assumes are not stated in the video.
  • Whether the email hit-rate lift (~30% → 80-90%) is Clay’s marketing figure or the creator’s measured result is ambiguous in the transcript.^[ambiguous]
  • Whether the Clay MCP will gain campaign/sequence management (closing the CSV-import handoff) — the stated roadmap gap.

Graph View

Backlinks