What it does
Long-running agent tasks execute asynchronously — user can close the tab, come back later, and find the work done.
Evidence
- Replit - Core Upgrade Modal — fifth Core feature.
- Icon: atom/orbital glyph.
- Related description in modal: "Build task plans in the background to get more work done at once."
Relationship to other Core features
- Replit - Autonomous Build Mode — the agent works unsupervised.
- Background Tasks — the work persists while the user is offline.
- Replit - Task Board — where the user tracks what's running.
- Replit - Plan Mode — what gets dispatched in the background.
💭
These 4 features together make a batch-oriented agent workflow. Plan (write the task list) → Dispatch (send to background) → Autonomous Execute (agent doesn't need prompts) → Return to Task Board later (see results). Totally different UX mental model than the chat-per-feature loop of the base tier.
Gating
Core-only. Free-tier users wait at the browser while the agent runs.
Differentiators vs Anything.com
- Anything.com - Max is autonomous but not explicitly "background/async" in any visible surface. Worth clipping Max's docs to verify.
- Background + Task Board together is a more project-management-flavored agent experience than Anything.com's chat-centric one.
Sources
- !raw/replit/Task_View.png