Super agents and sub agents, explained
The Coordinator
A super agent owns an entire business outcome. It decomposes the goal, assigns work to sub agents, mediates between them, handles errors and rollback, and escalates to a human when confidence drops.
The Specialists
Sub agents each do one thing well — ingest a document, verify an identity, score a risk, retrieve knowledge, observe a process. Seven archetypes ship out of the box: processor, analyzer, validator, aggregator, recommender, RAG, and observer.
Build the way your team works
Four development modes on one canvas — so business analysts and platform engineers ship the same workflow.
No-Code
Design on the visual canvas, or describe the workflow to the Ask AI conversational builder.
Low-Code
Configure agents through the onboarding UI without touching a repository.
Pro-Code
Write custom agent logic in Python against a published, versioned agent schema.
Hybrid
Mix marketplace agents and custom agents in the same super-agent workflow.
From draft to production, governed at every step
Orchestrations move through a controlled lifecycle, validated against the schema before they ever reach your catalog.
STEP 1
Draft
Start from one of 15 pre-built templates or a blank canvas.
STEP 2
Created
Assign sub agents and bind resources.
STEP 3
Configured
Set rules, guardrails, and human-handoff gates.
STEP 4
Tested
Run in a sandbox with batch evaluation.
STEP 5
Deployed
Promote through dev, QA, and production.
