Skip to main content
For work that splits into independent parts, Sixth can spawn 1 to 5 sub-agents that work at the same time in the same workspace. Each sub-agent takes one part. When all of them finish, Sixth gets their combined results and continues. For multi-part work this is much faster than doing each part in turn. You don’t need to set anything up: Sixth decides when to use sub-agents, using the spawn_agents tool. Sub-agents are different from Background Agents, which are specialists you define and that keep running while the main task continues. A sub-agent is temporary and Sixth waits for it to finish.

​
When Sixth uses them

Typical cases are tasks with clearly separate pieces, such as:
  • Adding the same kind of change across several unrelated modules or routes.
  • Building several independent components or pages.
  • Investigating different parts of a codebase at once.
Sixth gives each sub-agent detailed instructions (exact files, patterns to follow, how its part connects to the others) plus the project’s tech stack and conventions. Sub-agents know only what Sixth tells them.

​
What sub-agents can do

Each sub-agent can read, search and list files, read several files at once, write and edit files, create SVG or ICO files, delete files, run commands, and check background shells it started. When it’s done it reports a short result back to Sixth.

​
Following along in the chat

While sub-agents work, the chat shows a Sub-agents group with a summary such as “1 of 3 done”. Each sub-agent gets its own row with:
  • a status icon: a spinner while working, then done, failed or timed out
  • its name
  • what it’s doing right now
  • how long it has been running
  • a Stop button while it’s running
When a sub-agent finishes, click its row to see its result, the files it changed and its token usage.

​
Limits

  • At most 5 sub-agents run at the same time.
  • A single sub-agent can run for up to 4 hours.
  • Starting sub-agents is auto-approved whenever any auto-approve action is turned on.