Products / GTM Orchestrator

Production tested

The coordinator that runs your GTM pipeline.

A brief goes in. Napoleon, the coordinator, dispatches your crew through the as-run pipeline, discover through launch, and holds for your approval at the checkpoints that already fire on real campaigns.

Live

Three commands, the whole state of a campaign.

Every session starts with a read-only digest, never a guess from memory.

$ gtm-orchestrator status
queue: 2 decisions held - launch-gate, people-find-gate
$ tasks-axi ready
07-apply is the frontier task
$ campaigns show gtm-client-acme/hiring-q4
stage: 07-apply · as-run pipeline, discover through launch
Digest is CLI output, never a file read. Command names and stage name (07-apply) are the confirmed as-run CLI surface (AGENTS.md, CAMPAIGNS.md); example queue/stage above is illustrative, not a captured run. The wider 15-task campaign graph is built and tested but not yet exercised on a live campaign - see below.
PROOF NEEDED - a real captured session-start digest (actual held decisions, actual frontier task, actual stage count) from a live campaign, to replace the illustrative example above.
What it is

One coordinator. A crew it dispatches to.

GTM Orchestrator coordinates; it does not do the work itself. A brief moves through the as-run pipeline, and a fixed crew of specialist agents executes each stage under a brief with orient, task, guardrails, and return sections.

The as-run pipeline, discover to launch

Nine named stages run today - 01-discover through 09-launch, several with lettered sub-stages (04b-waterfall, 07b-personalize, 08-qa/mechanical). A wider 15-task campaign graph is built and tested, not yet exercised on a live campaign.

Gates that hold for you

Launch, paid people-find, and spend over the contract ceiling are confirm-gated. The operator's words are a claim, never proof - the coordinator verifies the result independently.

Checkpoints, not silent retries

A stage that needs a human decision writes a checkpoint file and stops. The operator edits it; the coordinator never fills one in for you.

Reads from the brain

Grounded in the GTM MCP - 1,626 campaigns of proven pattern, so its dispatch decisions start from real outcomes.

The crew it dispatches

Five specialists, one coordinator.

The roster is fixed - the coordinator never invents an agent. Each owner validates its own work; there is no separate QC pass.

  • List BuildingSourcing and filtering
  • EnrichmentCompany and people data
  • CopywritingSequences and personalization
  • EmailBisonInbox, list, and launch mechanics
  • CadenceSend timing and follow-up
How a campaign moves

Work is happening all the time. Approve next steps where your judgment matters.

01

Brief becomes a pipeline run

A client brief starts the as-run pipeline - discover through launch. A wider 15-task campaign graph with its own worktree and contract is built and tested, not yet exercised on a live campaign.

02

The coordinator dispatches

Each task goes to its owning specialist with an orient, task, guardrails, and return brief - never pasted context, never an invented agent.

03

It holds where judgment matters

Launch, paid people-find, and over-ceiling spend stop for your explicit yes. Everything else is auto or notify, never a stricter or looser gate than the table says.

04

State survives the restart

Task state, lead state, and cost state each live in exactly one place. Conversation memory is never authoritative - the backlog and the database are.

A thousand soldiers. One general. You.

One brief. A coordinated crew. Your approval at the gates that matter.

GTM Orchestrator ships with every founding seat, alongside the brain, the CLI, and a new production system every month.