Run pipeline init. Put your subscription to work.

get-strategy-snapshot

Preflight distillation of your GTM context into a compact, high-signal strategy snapshot: best ICP, commercial offerings, campaign offers, lead magnets, the only valid proof points (case studies), banned phrases, and the channels the Strategist may plan. Reason on this small snapshot instead of the raw GTM dump - fewer tokens, and any proof you cite traces to a real artifact so it survives validate-campaign-brief. Returns the snapshot object.

create-strategy-ideas

Run the Strategist diverge pass: brainstorm and curate campaign ideas (hook angle, archetype, personalization elements, CTA, proof point, differentiation, estimated audience) for your workspace. Same canonical engine the in-app Strategist uses, so the output matches the UI. Pass guidance when the operator has a specific ICP, offer, signal source, competitor post, or playbook direction. Returns { agentRunId, ideas }. Pick the strongest idea id and pass it to create-campaign-brief.

create-campaign-brief

Expand a chosen idea into a full structured campaign brief (ICP rubric, targeting filters, required research signals, message decision tree, frameworks, cadence). For kind=campaign pass the ideaId from create-strategy-ideas. kind=content or engagement build single-shot from workspace context (no ideaId). The brief persists as draft; run activate-brief to launch it. Returns { agentRunId, brief, kind }.

save-campaign-brief

Validate and persist a campaign brief authored by the caller. This deterministic write never invokes Pipeline hosted inference and always saves as a draft.

update-campaign-brief

Edit an existing campaign brief without hosted AI. Pass a partial plan, including name, frontEndOffer {name, description, whyIcpWantsIt}, or operatingContract with offerScore {deliveryCost, prospectValue, competenceProof, painFit, linkedinFriction, reasoning, total, verdict}. Scores are integers 1-5; total is their sum; verdict strong at 21+, usable at 16+, otherwise weak. Preserve other operatingContract fields. Asset selection uses selectedAssetBinding and is validated server-side.

preview-campaign-revision

Deterministically preview a prelaunch campaign revision. Returns changed fields, exact impact, row cohort, and a proposal fingerprint. Does not invoke AI or write data.

apply-campaign-revision

Apply an exact previewed prelaunch campaign revision after confirmation. Refuses stale previews and live or scheduled campaigns. Does not source, research, write, schedule, or send.

list-gtm-assets

List offering-scoped direct offers, front-end offers, and lead magnets, optionally filtered by lifecycle. Read-only.

create-gtm-asset-candidate

Create a caller-authored GTM asset candidate for one core offering. Does not invoke hosted AI or approve the asset.

update-gtm-asset-candidate

Edit an unapproved GTM asset candidate, or create a superseding version when the selected asset is immutable.

approve-gtm-asset

Approve and freeze one reviewed offering-scoped GTM asset version. Requires explicit confirmation.

retire-gtm-asset

Retire one GTM asset from new selection while preserving historical bindings. Requires explicit confirmation.

delete-gtm-asset-draft

Hard-delete only an unapproved, unbound GTM asset draft. Requires explicit confirmation.

validate-campaign-brief

Run the deterministic brief quality gate (the parity enforcer): proof points must trace to real GTM case studies/offers, decision-tree conditions must reference declared signals, the value proposition must link to a distinct ICP pain, and audience size must be sane. Pass briefId (a stored brief) OR plan (a raw brief plan you authored locally) - exactly one. mode launch (default) demotes generation-quality findings to warnings; strict treats them as errors. Returns { valid, errors, warnings }. Always run this before activating a brief you reasoned yourself.

activate-brief

Run the brief: validate it (launch-mode gate), flip it to active, and provision its agent-pipeline enrichment table. scoutExecutionMode defaults to manual, which pauses before hosted Scout work. pipeline_hosted enqueues Scout and requires allowHostedAi=true. A brief that fails always-broken validation is refused (422) unless force=true. Returns { activated, briefId, status, alreadyActive, scoutEnqueued, scoutPaused, scoutExecutionMode, warnings }.