# Specialists can run long commands in the background

> A specialist can now start a long build, scrape, or render, keep talking to you, and pick the result up on its own when the command finishes.

Nov 6, 2026. Topics: Agents, Platform. https://www.generalinput.com/changelog/2026-11-06-background-commands

A specialist no longer has to sit and wait for a slow command. It can start a long build, a data pull, or a video render in the background and keep working with you in the same conversation. When the command finishes, the specialist picks up on its own and tells you how it went.

You can see it working. The conversation shows as busy in your lists, and each command sits in the chat as a card showing the latest line it printed, then stays there with how it ended. On the web, the sidebar's Running list shows what is running, where you can watch a command's output and stop it. On the phone, tap the card to do the same, or tap the specialist's name for a page with everything running and the files it is working with.

### How it works

- **Nothing gets cut off.** A command that runs longer than expected moves to the background by itself instead of dying at a time limit, so the specialist never has to guess how long something will take.
- **The same access as any command.** It runs in the conversation's own sandbox with the connections and [Agent Firewall](/agent-firewall) rules of a normal one, and passwords and tokens are removed from its output as it is written.
- **Up to three at once.** Each conversation can run three background commands, each for up to an hour.
- **The specialist reports back.** It takes a turn when a command ends. If you are mid-conversation, the report waits.
- **Billed as sandbox time.** The sandbox stays up while a command runs, at the usual 80 credits an hour.

### Also shipped

- If a sandbox goes away mid-command, the conversation is told the command was lost rather than left waiting.