Source
● YC · S26
ay@marops:~$ ./run lead-scoring --source yc-s26 ▋
Unit MO-1 / Lead Scoring ICP Lead
ICP Lead
Scoring
Ingests a public company list, scores every entry against one ICP, and outputs a ranked outbound queue. No manual triage. Click a row for the grounded rationale, signal, and recommended play.
AOUT
BNURT
CSUPP
Model
claude -p
Scored
8/25
Tiers
7 A · 7 B · 11 C
Generated
2026-06-11
ICP ▸ Seed / early-stage B2B companies deploying LLM agents to production — the buyer for an AI agent observability and evaluation platform.
CLAY ▸ The same ICP logic rebuilt no-code in Clay — Claygent research, an AI scoring column, formula-derived tiers, and a HubSpot write-back. Walkthrough:
Embed blocked on your network? Watch on Loom →
N8N ▸ And a third time as a low-code n8n workflow — same loop in code, no-code, and low-code, because the logic matters more than the tool. Claude proposes the score; a deterministic Switch node disposes the tier (the 80/55 cutoffs are branches on the canvas, not model output). Run on the same 25 real companies.
01
Companies
25
scored this run
02
A-Tier ≥80
7
routed to outbound
03
Median
62
full batch
04
Cost / 1K
$1.60
est. at token profile
Filter / search
Tier
# Company ICP Score ▾ Tier
FIG.1 — ICP scoring / rev.A — Data: real companies, Y Combinator Spring 2026 batch (public directory). One-liners are the companies' own. Scores are model-generated against the stated ICP, on a sample run.