Founder operations

GTM Agent Swarm

The system is deliberately approval-gated: outputs can be drafted and reviewed, but not sent, published, scraped, or enriched automatically.

Focus

Approval-gated GTM workflow

PythonSQLiteAgent orchestration

GTM Agent Swarm

DraftQueueReplay

Awaiting approval5

Runs traced142

Avg cost / run$0.034

Autonomous sends0

Command lane

› draft:post launch-teaserqueued

› draft:reply pricing-threadqueued

› memo:strategy q3-positioningqueued

› eval:trace run-0142report

› replay post-017match

drafts route to the approval queue — nothing leaves it

›

Approval queue

Post

Launch teaser — draft v3

Awaiting approval

Pricing thread response

Awaiting approval

Memo

Q3 positioning memo

Awaiting approval

Post

Founder build-log #12

Awaiting approval

Integration question

Awaiting approval

Send: disabled · Publish: disabled · Enrichment: disabledScrape: disabled · CRM: disabled · Outbound webhooks: disabled

Trace / cost eval

run-0142Deep tier$0.062

run-0141Mid tier$0.031

run-0140Fast tier$0.009

Artifact replay

memo-006post-016reply-021post-017

Deterministic replay4/4 match

Sample data

Interactive rendition · sample data

The challenge

Founder GTM work benefits from repeatable drafts and traceability without outsourcing judgment or external action to an agent.

The approach

Build typed contracts, model routing, queue states, and traceability around a review-only workflow.

01
Content, strategy, and reply drafting
02
Approval-gated queue and artifact replay
03
Cost, trace, and routing evaluation
04
Founder-supplied campaign workflow

Keep human review and explicit delivery boundaries central as the workflow grows.

Discuss this project