Build GTM Agents with Claude

What is a GTM agent?

An autonomous agent that monitors your PostHog metrics, identifies when plays need adjustment, and takes action — updating sequences, pausing underperformers, or escalating to the founder.

Architecture

1. Trigger: n8n cron job runs every 24h (or on PostHog alert)
2. Gather: Agent pulls current metrics from PostHog MCP, deal status from Attio MCP
3. Analyze: Claude evaluates metrics against play thresholds and the 20% guardrail rule
4. Decide: Agent chooses: continue, adjust, pause, or escalate
5. Act: Agent executes the decision via n8n (update sequences, send alerts, modify campaigns)
6. Report: Agent logs its decision and reasoning to a shared channel

Guardrails

• Always require human approval for: budget increases >20%, new play launches, pipeline deletions
• Log every agent decision with reasoning for audit
• Set hard stops: if conversion drops >40%, pause all automation and alert the founder