Changelog

Sub-agents

Adapt now breaks complex requests into parallel workstreams handled by specialized agents.

Some questions can't be answered with a single query. They require pulling data from multiple sources, running analyses in parallel, synthesizing conflicting signals, and presenting a unified answer. Sub-agents make this possible.

How it works

When Adapt receives a complex request, it decomposes it into discrete sub-tasks and delegates each to a specialized agent. These agents work in parallel, and their results are synthesized into a single coherent response.

You ask one question. Adapt orchestrates everything behind the scenes.

Example:

Give me a full picture of why churn spiked in March.
Pull usage data, support tickets, and any relevant Slack discussions.

Adapt might spawn three sub-agents simultaneously:

  • One querying your analytics platform for usage drop-off patterns
  • One searching support tickets for themes during that period
  • One scanning Slack for customer mentions and internal discussions

Results come back together, cross-referenced and synthesized.

Why this matters

Previously, multi-source analyses required you to run several separate queries and manually connect the dots. Sub-agents collapse that into a single interaction. You get a complete answer without having to orchestrate the research yourself.

This is particularly useful for:

  • Root cause investigations across product, support, and usage data
  • Competitive research pulling from web, CRM, and internal notes simultaneously
  • Executive briefings that need context from engineering, finance, and sales data in one place
  • Incident retrospectives correlating events across multiple systems

Transparency

Adapt shows you what each sub-agent was tasked with and what it found. You can expand any sub-agent's work to see the raw query, the sources it used, and its intermediate conclusions before the final synthesis. Nothing is hidden.

Limits

Sub-agent tasks run within your existing integration permissions. An agent won't access a data source you haven't connected, and it won't take actions you haven't enabled. The same authorization model applies whether the request comes directly from you or from an orchestrating agent.

Get started with up to $300 in credits for you and your team.