Schedule Claude Code tasks on the web
Claude Code just rolled out web-based scheduled tasks, enabling users to automate recurring prompts for code reviews, security audits, and more, all within Anthropic's cloud infrastructure. This feature immediately resonated with developers keen on AI-driven software iteration, seeing it as a significant step towards fully automated dev workflows. However, the announcement quickly spurred debate on its cost model, usage limits, and the perceived lack of transparency from Anthropic regarding these details.
The Lowdown
Anthropic's Claude Code has launched a new feature that allows users to schedule AI tasks directly on the web, leveraging Anthropic-managed cloud infrastructure. This capability ensures that AI prompts can run on a recurring basis, even when the user's local machine is off, providing a powerful tool for continuous code-related automation.
- Core Functionality: Automates recurring AI prompts for development tasks such as reviewing pull requests, analyzing CI failures, syncing documentation, and running dependency audits.
- Availability: Accessible to all Claude Code on the web users, including Pro, Max, Team, and Enterprise plans.
- Scheduling Flexibility: Offers three methods for scheduling: Cloud (Anthropic cloud, no machine required, persistent), Desktop (user's machine, local file access, persistent), and
/loop(CLI session-scoped). - Task Creation: Scheduled tasks can be set up via the web interface, the Desktop app (for remote tasks), or through the CLI using the
/schedulecommand. - Frequency Options: Provides preset frequencies (Hourly, Daily, Weekdays, Weekly) with automatic time zone conversion, and allows for custom intervals via CLI updates.
- Repository Integration: Tasks clone specified repositories on each run and, by default, restrict AI pushes to branches prefixed with
claude/, with an option to remove this restriction. - External Service Connectivity: Supports MCP connectors, enabling tasks to interact with external services like Slack or Linear for reading and writing data during execution.
- Configurable Environments: Each task runs within a configurable cloud environment, controlling network access, environment variables, and setup scripts.
- Management: Tasks can be viewed, edited, paused, resumed, or deleted from a dedicated web page or using CLI commands such as
/schedule list,update, andrun. Past runs are stored as full sessions for review and interaction.
This new cloud-based scheduling capability aims to deeply integrate AI into developer workflows, promising enhanced efficiency and continuous monitoring of codebases without requiring constant manual intervention.
The Gossip
Automating Away Anxiety: AI's Agile Assist
Users enthusiastically discussed the practical applications of Claude Code's scheduled tasks, seeing immense potential to automate tedious, recurring development tasks. Examples included automated security audits, in-depth code reviews, and continuous error monitoring. Many commenters envisioned a future of rapid software iteration where AI agents handle everything from feedback analysis to code deployment, marking a significant step towards a fully AI-driven development lifecycle.
Costly Clauses and Cloud Conundrums
A significant portion of the discussion revolved around the pricing model, usage limitations, and Anthropic's transparency. Users expressed frustration over the lack of clear pricing for compute resources, the perceived vagueness regarding token costs, and the unexpected imposition of scheduled task limits (e.g., 3 tasks) even for high-tier subscription plans. The discovery of these restrictions through social media rather than official documentation also fueled accusations of a 'rugpull' by some users.