Artifact Creation
The user explicitly asks for a chart, diagram, dashboard, sankey, architecture diagram, html report, kpi grid, table, or any other visual artifact
Agent Types
Built-in sub-agent personas the supervisor can delegate to, source-derived from the API built-in agent directory.
Built-in Agents
8 built-in agent types shipped with the release
Cloud Operations
Multi-step cloud audits, cross-cloud comparisons, troubleshooting workflows, and reasoning over enumeration results that need chained reads (e.g
Code Execution
The user asks you to write, run, or debug a script, transform data with code, run a quick numerical computation
Data Query
The user asks to query, retrieve, or aggregate structured data from a known datastore the platform manages — Postgres, Milvus (vector retrieval), Athena/Redshift, BigQuery — and…
Planning
The user asks for a multi-step plan — a runbook, a migration sequence, an incident-response playbook, a project breakdown, or any ordered set of actions where the order, owners,…
Reasoning
The user's question requires deep, multi-step thinking that the supervisor's primary-loop turn cannot solve in one shot — long-horizon planning, novel problem decomposition, mat…
Synthesis
The supervisor has gathered evidence from several tool calls or sub-agents and needs a focused, well-organised final answer composed for the user — turning raw findings into pro…
Validation
The supervisor needs an independent second look at a tool result, a draft answer, or a completed sub-task to confirm it's correct, complete, and free of hallucination before the…