Skip to content
Auteryn

Background & Long-Horizon Execution

Not every task fits in a single synchronous reply. Auteryn agents can take on long, autonomous work that runs in the background — kicked off, worked on over time, and picked back up if interrupted — instead of blocking on one request/response turn.

Start a task, walk away, and come back to the result. Background runs execute asynchronously, can schedule their own follow-ups, and are resilient to interruptions.

What “long-horizon” means

Section titled “What “long-horizon” means”

Asynchronous runs

Queue a task to run in the background and get results when it’s done — no need to keep a connection open.

Self-scheduled follow-ups

An Internal agent on a paid plan can schedule a follow-up run for later, to check back on something or continue after a wait.

Resilience & resume

If a long run is interrupted, the agent can resume with its prior plan, recent results, and working notes intact.

Beyond one reply

Multi-step research, coding, and workflow tasks that span far more than a single chat turn.

Running a task in the background

Section titled “Running a task in the background”

You can start a background run in two ways:

  • In the workspace — kick off a Deep task and let it run while you do other things; results stream into the conversation as they complete.
  • Via the API — submit a run asynchronously, then check on its progress and read the result.

POST /api/run/async returns 202 with the run and task IDs at once while execution continues server-side. Check progress with GET /api/run/{task_id}/status, or stream it with GET /api/tasks/{task_id}/stream. See Async & background runs.

Recovery & continuation

Section titled “Recovery & continuation”

If a long run is interrupted (a timeout, a cancellation or a restart) or fails, the next message can include recovery context, which gives the agent back its previous plan and notes along with the results of its last few steps. This is within-conversation continuation, so the agent doesn’t start over from scratch. See Autonomous execution for execution modes and timeouts.

When to use it

Section titled “When to use it”
  • Deep research that reads and synthesizes many sources
  • Multi-step coding in a sandbox
  • Scheduled or recurring work driven by Loops
  • Anything that takes longer than a quick, interactive reply
Section titled “Related”