Claude-first GTM motion - anyone built this?

emmareach

Been trying to build a full GTM engineering motion around Claude, and it's a bit of a minefield. The idea is to use Claude as the core operating layer - not just for copy, but for actually running workflows: ICP research, lead discovery, account qualification, enrichment, personalised outreach, campaign creation, CRM updates, follow-ups, and reporting. Everything, basically.

What I'm really after is whether there are decent ready-made GitHub libraries, MCP servers, or agent frameworks that work out of the box. I've seen some setups where people use Claude as orchestration with MCP integrations for HubSpot, Slack, and Notion. A colleague swears by a flow using /last30days for research and then Claude/Codex for everything else. Others warn against fully autonomous SDR agents - better to build composable workflows with human review layers. That makes sense for reliability.

anyone else tried this? Keen to hear what's actually working in the wild.

neonlights

Yes the ecosystem around Claude based GTM engineering has grown fast recently, especially around MCP servers, agent frameworks, and workflow automation. Most strong setups use Claude as the orchestration layer for ICP research, enrichment, personalization, CRM syncing, and reporting, combined with MCP integrations for tools like HubSpot, Slack, Notion, and GitHub. The teams seeing the best results usually avoid fully autonomous SDR agents and instead build smaller composable workflows with human review layers, because that tends to be far more reliable and scalable in real-world outbound operations

logan-lifecycle

here are some good public repos
https://github.com/growthenginenowoslawski/coldoutboundskills/
https://github.com/kenny589/claude-code-gtm-starter-kit
https://github.com/ivangfalco/gtm-skills

harper-local

Relying entirely on Claude for GTM just sounds like a recipe for a massive, unscalable mess of prompt drift.

dusting

Claude handles the bulk of the heavy lifting internally-research, pattern recognition, sifting through public data for ICP signals. For the rest, I lean on Codex alongside a CLI-native toolchain that ties GTM integrations and data providers together. It's about reducing friction between insight and action, less about choosing tools and more about designing a workflow that lets you move from signal to send in one fluid motion.