Claude Code Routines: Automated Cloud Tasks for AI Development Workflows

Claude Code Routines are saved configurations that package a prompt, one or more repositories, and connectors into automated tasks that execute on Anthropic-managed cloud infrastructure. This means routines continue running even when your local machine is offline.
How Routines Work
Each routine can have multiple trigger types attached:
- Scheduled triggers: Run on recurring cadences like hourly, nightly, or weekly
- API triggers: Execute on demand via HTTP POST to a per-routine endpoint with bearer token authentication
- GitHub triggers: Run automatically in response to repository events including pull requests, pushes, issues, or workflow runs
A single routine can combine multiple triggers. For example, a PR review routine could run nightly, trigger from a deploy script via API, and also react to every new pull request.
Practical Use Cases
The documentation provides several concrete examples of routine applications:
- Backlog maintenance: A schedule trigger runs nightly against your issue tracker via connector, reads new issues, applies labels, assigns owners based on code areas, and posts summaries to Slack
- Alert triage: Monitoring tools call the routine's API endpoint when error thresholds are crossed, passing alert bodies as text. The routine pulls stack traces, correlates with recent commits, and opens draft PRs with proposed fixes
- Bespoke code review: GitHub triggers on pull_request.opened apply team-specific review checklists, leave inline comments for security/performance/style issues, and add summary comments
- Deploy verification: CD pipelines call the routine's API endpoint after production deploys. The routine runs smoke checks, scans error logs for regressions, and posts go/no-go decisions to release channels
- Docs drift detection: Weekly schedule triggers scan merged PRs, flag documentation referencing changed APIs, and open update PRs against docs repositories
- Library porting: GitHub triggers on pull_request.closed (filtered to merged PRs) port changes to parallel SDKs in other languages and open matching PRs
Creating and Managing Routines
Routines are available on Pro, Max, Team, and Enterprise plans with Claude Code on the web enabled. You can create and manage them at claude.ai/code/routines or from the CLI with the /schedule command.
Creation options include:
- Web interface at claude.ai/code/routines
- Desktop app: Click "New task" and choose "New remote task" ("New local task" creates local Desktop scheduled tasks instead)
- CLI with
/schedulecommand
The creation form configures the routine's prompt, repositories, environment, connectors, and triggers. All creation methods write to the same cloud account, so routines created via CLI appear immediately in the web interface.
Current Status and Limitations
Routines are currently in research preview, meaning behavior, limits, and the API surface may change. They execute as full Claude Code cloud sessions without permission-mode prompts.
📖 Read the full source: HN AI Agents
👀 See Also

SquarePact Word Add-In Connects to OpenClaw Gateway for Agentic Document Editing
SquarePact's Word add-in connects to your own OpenClaw gateway over websocket. The agent proposes edits shown as approval cards, with stale edit detection and retry logic for malformed JSON from smaller models.

bad-ass-mcp: Free, Open Source MCP for Native Desktop GUI Control via Accessibility API
bad-ass-mcp is an open source MCP server that lets Claude and other AI agents control macOS, Windows, and Linux desktops using the native accessibility layer — no screenshots, no look-move-look loops. Free alternative to Computer Use, Operator, or UiPath.

ClawCall Adds Inbound Calling — Your OpenClaw Agent Gets Its Own Number
ClawCall now supports inbound calling — your OpenClaw agent gets a dedicated number, picks up within seconds, and talks with your tone. Easier greeting and personality setup via natural language.

Claude Code vs. Codex: Real-World Build Test – 36 Files vs. 28, Infinite Loop, and $0.46 Cost Difference
A developer pits Claude Code against Cursor's Codex on two real tasks: a PR triage bot and a WebSocket code review UI. Claude built 36 files in 12 minutes with zero TypeScript errors; Codex produced a working UI but hit an infinite React loop. Cost difference: ~$0.46.