Orchestrun.AI

Platform

A team of agents, each doing one job well

Orchestrun doesn’t bolt a single chatbot onto your marketing stack. Blueprint, Canvas, Pulse, and Orchestrator each own a distinct stage of the campaign lifecycle — with their own tool-calling systems, memory, and approval gates.

Blueprint

Strategy & setup agent

Guides campaign creation through a structured, multi-step conversational intake — audience captured as structured data, not freeform text.

  • Auto-detects the user's language at conversation start and locks to it unless asked to switch.
  • Produces a frozen campaign-plan snapshot that flips the campaign into a pending-review state.
  • A named human must approve the plan — the creator cannot self-approve (segregation of duties).
  • Can promote an approved plan directly into a multi-step customer journey via Orchestrator.
  • Chat sessions persist server-side for 24 hours, so a conversation survives a page reload.

Canvas

Creative agent

An 8-step creative flow that generates channel-specific text variants — tone and length tuned per channel.

  • Auto-repurposing: one piece of text content fans out into LinkedIn, Instagram, and WhatsApp-ready variants.
  • Pluggable image-generation provider architecture, ready to connect to a model provider.
  • Every agent-generated asset goes through a real approval workflow before it can be used.
  • Content Studio: a real asset library with review status, filtering, and campaign linking.

Pulse

Performance agent

Real-time metrics ingestion and a threshold-driven optimization engine.

  • Proposes creative swaps, channel fallback, budget reallocation, and A/B test winner selection.
  • Every optimization proposal requires human approval, applied transactionally.
  • Full A/B test orchestration with automated winner evaluation.
  • Holds a narrowly-scoped write capability — proposing budget reallocation between live paid channels — spend-cap enforced and gated by a dedicated approval permission.
  • Can auto-approve certain reallocation proposals once an account has earned autonomy trust for that gate, with full audit reasoning logged.

Orchestrator

Journey agent

Turns an approved campaign plan into a visual, drag-and-drop customer journey.

  • 5 node types — Send, Delay, Branch, A/B Split, Fallback — on a real graph engine with cycle and unreachable-node validation.
  • Journey versioning: every edit creates an immutable snapshot; running participants keep following the version they started on.
  • Preview / fast-forward test mode runs a real send through the full consent, suppression, and quiet-hours pipeline, compressed into seconds.
  • Can be gated behind the same human-approval-or-autonomy framework as campaigns and budgets.

Human in the loop

AI moves fast. You stay in control.

Segregation of duties

A campaign's creator cannot approve their own campaign. The same block applies to asset approval and journey-activation approval — enforced structurally, not by convention.

Human-gated by default

Every consequential agent action — plan, creative, budget move, journey activation — starts fully human-gated. Nothing auto-executes on day one.

Autonomy is opt-in and reversible

Accounts build measurable trust per gate type. Once trust crosses an admin-set threshold, low-risk actions can auto-approve — with a one-click kill switch and full audit reasoning on every autonomous decision.

Model flexibility

Bring your own model

Blueprint, Canvas, and Pulse can each be independently configured to run on OpenAI, Anthropic, Azure OpenAI, or a self-hosted model — so you’re never locked into a single LLM vendor.