A complete GTM
Operating System
Every job RevOps does — quota planning, lead routing, deal health, churn risk, customer comms, business reviews — specced as a service. Three reasoning agents on top read those services to answer real CRO questions. 41 services + 3 agents + 15 specialist tools across 10 layers.
Three tiers. 41 GTM Services in L1–L8 (deterministic, single-writer-per-table, audit-grade — v38 added AGT-208 Developer Signal Scorer for PLG-funnel scoring on consumption-pricing platforms). 3 Brain Agents in L9 (LLM-native, operator-invoked, read Tier 1 services but never write canonical data; gated by 30-question eval harness — all three prototyped & eval-passing — AGT-903 added v38.6 with 5/5 EVAL-S fixtures green at \$2.89 sweep). 15 Specialist Tools in Tier 3 across 4 waves — stateless callable functions that brains and services invoke for narrow cognition tasks (TOOL-003 / TOOL-004 / TOOL-008 / TOOL-010 / TOOL-013 / TOOL-014 / TOOL-015 prototyped; remaining 8 specced). The distinction is deliberate; brains propose, humans co-define and approve, services execute. v38 also adds tier-migration as a new expansion play type (AGT-503), GP-overlay comp variant (AGT-101 + AGT-103), and motion-aware lead routing (AGT-202). See changelog for current state.
Service Connection Map
Deterministic JS reproduction of AGT-201 6-dimension scoring + AGT-202 routing. Useful for explaining the dimensional model — narrow scope, no LLM, no brain reasoning. For end-to-end reasoning see Brain Outputs and Follow a Lead.
eval/samples/brain_outputs.json…