Robots Atlas>ROBOTS ATLAS
Agents

Background Agents

2024ActivePublished: 6 August 2026Updated: 6 August 2026Published
Key innovation
A shift from interactive, step-by-step steering of an agent to delegating whole tasks to agents running asynchronously in the background - without blocking the user and without confirming every action.
Category
Agents
Abstraction level
Pattern
Operation level
OrchestrationApplicationTooling
Use cases
Autonomous coding tasks (code, tests, pull requests)Running many long-running tasks in parallelRepository-scale refactoring and migrationsBackground workflow automationLong-horizon research and analysis tasks

How it works

The user delegates a task (e.g. 'implement feature X' or 'refactor module Y'). The background agent starts asynchronously - locally or on a remote machine - and runs a loop: planning, calling tools (editing files, running tests, commands), evaluating and iterating, until completion. State is often persisted in an event log (every model call, tool run and edit), enabling long-running tasks and crash recovery. The result is delivered as a finished artifact (e.g. a pull request, a report), and many agents can run in parallel.

Problem solved

Interactive agents require constant user attention and confirmations at each step, which blocks the user's work and limits scale (one task at a time). This makes it hard to delegate large, long-running tasks.

Components

Task queue / delegationWork intake and scheduling.

A mechanism for accepting whole tasks from the user and starting them asynchronously.

Execution environment (VM / local)Where the task executes.

An isolated environment (often a remote VM) where the agent runs tools without loading the user.

Official

Persistent event log / statePersistence and resilience.

A record of every model call, tool run and edit, enabling long-running tasks and crash recovery.

Artifact deliveryThe pattern's output.

Returning a finished result (e.g. a pull request, a report) once the task is complete.

Official

Implementation

Implementation pitfalls
Acting without confirmationHigh

Not confirming each step can lead to unwanted changes if the task was poorly specified.

Fix:Approval gates for risky actions, isolated environments and review of the artifact (e.g. a PR) before merging.
Harder observabilityMedium

Background tasks are less visible to the user, making progress tracking and debugging harder.

Fix:A persistent event log, status visibility and clear final reports.

Evolution

2024
Cloud autonomous coding agents (Devin)
Inflection point

Cognition introduced Devin - an autonomous agent executing whole coding tasks in the cloud, popularizing the 'delegate a task' approach.

2025
Background Agents in developer tools

Cursor, GitHub Copilot and others introduced background agents run asynchronously (often in the cloud), returning results as pull requests.

2026
Persistent per-session background agents (Meta Muse Code)

Muse Code introduced specialized background agents that remain active throughout a session, with a local event log and repository-scale execution.