All capabilities Agents, tools and MCP

Hand off a task and walk away from the machine

Start the work in the terminal, move it to the cloud, check it from a phone, and have it run again next Tuesday without you.

DocsOfficially documentedTestNot tested by benchr Medium · Agent / CLI Docs checked September 1, 2026

Claude CodeClaude Opus 5Claude Sonnet 5

Overview

Runs sessions in the browser with no local setup, so a long task keeps going when the laptop closes.
Moves a session between surfaces - terminal to web, web back to terminal, terminal to desktop - against the same engine and project settings.
Runs on a schedule: cloud routines that survive a powered-off machine, desktop scheduled tasks with local file access, and an in-session loop for quick polling.
Can be triggered from a team chat mention or a GitHub event rather than by a person at a keyboard.

01Write the task as an outcome with a verification step - the agent needs to know when it is finished, not just what to do.
02Pick where it runs by what it needs to touch: cloud for repositories, desktop scheduling for local files.
03Give it the project's rules in CLAUDE.md so an unattended run does not invent conventions.
04Review the branch, not the transcript. Unattended work is judged by its diff.

Cloud and mobile handoffs require a claude.ai subscription, and some handoffs are limited to macOS and x64 Windows.
An unattended agent is only as safe as the permissions and hooks around it.
Scheduled runs still cost tokens on every run, whether or not there was anything to do.

benchr has not run this capability. No test has been published.

benchr test runs →

Source-check history

September 1, 2026DocsOfficially documentedFirst logged. Web sessions, routines, desktop scheduled tasks, chat and GitHub triggers, and the subscription requirement all read from the product overview.