icp-agent

Autonomous lead-engagement agent. Receives {email, domain}, enriches, scores against an ICP, routes Hot/Warm/Cold, drafts a personalized first email, logs every decision.

Live status

Endpoint
live https://icp-agent-ten.vercel.app
LLM
checking…
Enrichment
checking…
Decision log
checking…

Try it

curl -X POST https://icp-agent-ten.vercel.app/api/webhook/inbound \
  -H "content-type: application/json" \
  -d '{"email":"founders@linear.app","domain":"linear.app"}'
Returns 202 {lead_id, sheet_row_url} immediately. The background agent runs ~15s and writes a fully scored row to the decision log.

What happens to each lead

POST /api/webhook/inbound {email, domain} │ ▼ validate + idempotency │ append placeholder row → Sheet │ respond 202 {lead_id} ← returns immediately │ waitUntil(run agent) ← background │ ┌──── enrich (parallel) ────┐ │ scrape(domain) [cheerio] │ │ hunter.domainSearch │ └────────┬───────────────────┘ │ ┌──── LLM assess loop (bounded ReAct, ≤2 deepen rounds) ────┐ │ action ∈ { score_now, fetch_linkedin, fetch_news, │ │ fetch_email_finder } │ │ agent picks; code executes; context grows; loop. │ └────────┬───────────────────────────────────────────────────┘ ▼ LLM final score → {score 0-10, criteria_breakdown, reasoning} ▼ route to Hot ≥8 / Warm 4-7.99 / Cold <4 → LLM email ▼ Sheet row → status=done with full decision chain

Endpoints

Links

GitHub repo Decision log (Google Sheet) /api/health