Schedule tasks on the web

A scheduled task runs a signal on a recurring rhythm using anthropically-managed infrastructure. Tasks continue to work even when your computer is turned off.
Some examples of recurring tasks you can automate:

  • Reviewing open pull requests each morning
  • Analyzing and summarizing overnight CI failures
  • Syncing documents after PR merge
  • Running dependency audits every week

Scheduled tasks are available for all Cloud Codes on web users, including Pro, Max, Team, and Enterprise.

Compare Scheduling Options

Cloud Code provides three ways to schedule recurring tasks:

Cloud desktop /loop
keeps on going anthropomorphic cloud your machine your machine
need to start the machine No Yes Yes
Open session required No No Yes
continuously during restart Yes Yes no (session-area)
access local files No (Fresh Clone) Yes Yes
mcp server Connectors configured per task config files and connectors Inherited from session
prompts permission No (runs autonomously) Configurable per task Inherited from session
Customizable Schedule through /schedule in CLI Yes Yes
minimum interval 1 hour 1 minute 1 minute
Use cloud work For work that must run reliably even without your machine. Use desktop work When you need access to local files and tools. Use /loop For quick voting during a session.

Create a Scheduled Task

You can create a scheduled task from three places:

  • web: Go to claude.ai/code/scheduled and click new scheduled task
  • desktop app:open Schedule page, click new workchoose more new remote work. See Desktop scheduled tasks for details.
  • CLI: run /schedule In any session. The cloud walks you through the setup via conversation. You can also pass a description directly, like /schedule daily PR review at 9am.

The web and desktop entry points open a form. The CLI collects the same information through guided interactions.
The steps below are run through the web interface.

Name the task and write the prompt

Give the task a descriptive name and write down the prompt for Claude Run each time. The prompt is the most important part: the task runs autonomously, so the prompt should be self-contained and clear about what needs to be done and what success looks like.Prompt input includes a model selector. The cloud uses this model for every task.

Select Repository

Add one or more GitHub repositories for the cloud to work. Each repository is cloned at the beginning of the run, starting from the default branch. creates clouds claude/-Prefix branches to its transformations. To allow pushes to any branch, enable Allow unrestricted branch pushes For that store.

Choose an environment

Select a cloud environment for the work. The environment controls who has access to the cloud session:

  • network access: Set the level of Internet access available during each run
  • Environment Variables: Provide an API key, token, or other secret that the cloud can use
  • setup script: Run install commands before each session starts, such as installing dependencies or configuring tools.

A default The environment is available out of the box. To use a custom environment, create a custom environment before creating the job.

choose a schedule

Choose how often the task runs from the Frequency options. The default is 9:00 am daily in your local time zone. Due to faltering, tasks may run a few minutes behind their scheduled time.If the preset options do not suit your needs, choose the closest option and update the schedule from the CLI /schedule update To set a specific schedule.

review connectors

All your connected MCP connectors are included by default. Remove everything that is not needed for the task. Connectors give the cloud access to external services like Slack, Linear, or Google Drive during each run.

create task

Click create. The task appears in the Scheduled Tasks list and runs automatically at the next scheduled time. Each run creates a new session with your other sessions, where you can see what the cloud did, review changes, and create a pull request. To trigger an immediate run, click run now From the job details page.

frequency option

The schedule picker provides preset frequencies that handle time zone conversions for you. Choose a time in your local area and the job runs at that wall-clock time, no matter where the cloud infrastructure is located.

Tasks may run a few minutes later than their scheduled time. The offset is consistent for each task.

frequency Description
per hour Runs every hour.
daily Runs once daily at the time you specify. The default local time is 9:00 am.
working days Similar to Daily but skips Saturday and Sunday.
weekly Runs once per week on the day and time you specify.

For custom intervals like every 2 hours or first of every month, choose the nearest preset and update the schedule from the CLI /schedule update To set a specific schedule.

Repository and branch permissions

Every repository you add is cloned on every run. Starts from the default branch of the cloud repository unless your prompt specifies otherwise.
By default, the cloud can only push to prefixed branches claude/. This prevents scheduled tasks from accidentally modifying protected or long-lived branches.
To remove this restriction for a specific repository, enable Allow unrestricted branch pushes For that repository when creating or editing a task.

connectors

Scheduled tasks can use your connected MCP connector to read and write from external services during each run. For example, a task that tests support requests might read from a Slack channel and cause issues in Linear.
When you create a task, all your currently connected connectors are included by default. Remove anything that isn’t necessary to limit access to the cloud during run. You can also add a connector directly from the task form.
To manage or add connectors outside the task form, go to Settings > Connectors at claude.ai or use /schedule update In CLI.

environment

Each task runs in a cloud environment that controls network access, environment variables, and setup scripts. Configure the environment before creating the task to provide access to the cloud API, install dependencies, or restrict the network scope. See Cloud environments for a complete setup guide.

Manage scheduled tasks

Click on a task in scheduled To open the list’s details page. The Details page shows a list of the job’s repository, connector, prompt, schedule, and previous runs.

View runs and interact with them

Click on any run to open it as a full session. From there you can see what the cloud did, review changes, create a pull request, or continue the conversation. Each run session works like any other session: use the dropdown menu next to the session title to rename, archive, or delete.

Edit and control tasks

From the Job Details page you can:

  • Click run now Starting the race immediately without waiting for the next scheduled time.
  • Use toggle in rebroadcast Section to pause or resume the schedule. Stopped tasks retain their configuration but do not run until you re-enable them.
  • Click the edit icon to change the name, prompt, schedule, repository, environment, or connectors.
  • Click the delete icon to delete the task. Previous sessions created by the task remain in your session list.

You can also manage tasks from the CLI /schedule. run /schedule list To see all the work, /schedule update to change any work, or /schedule run To trigger someone immediately.



<a href

Leave a Comment