Learn how Claude routines work.
Short guides in plain words. They cover Claude scheduled tasks (also called routines), connectors, artifacts, Claude Code plugins and agentic workflows (jobs where Claude makes judgment calls inside a process). Each one says which claims come from Anthropic's docs and which we checked by running them.
Nine guides, in the order to read them
Start with the first if Claude scheduled tasks are new to you. The rest stand alone.
What are Claude routines?
A saved Claude Code setup that runs on Anthropic's cloud on a schedule. What it can do and where it stops.
ReadSchedule Claude Code
Run Claude Code every day without a laptop open. Cron is in UTC and the shortest gap is one hour.
ReadClaude connectors
How a routine reaches mail, calendars and other services, and why every write needs a gate.
ReadClaude artifacts as apps
An artifact with a database can hold a routine's results. The Desk is one worked example.
ReadClaude Code plugins
How to install a plugin from a marketplace, and what the RoutineOS plugin contains.
ReadAgentic workflows
What to leave to code, what to leave to Claude, and how to keep it safe: report first, ask before acting.
ReadHuman in the loop AI
Approval gates for routines: what asks you first, and what a reply can and cannot do.
OpenClaude MCP servers
What an MCP server is, with a free one to add today.
OpenClaude skills on a schedule
How a skill is delivered to a scheduled run, and which routes are verified.
OpenWhere each fact comes from
| Tag | Means |
|---|---|
| pilot-verified | A real run checked it. |
| docs-stated | It comes from Anthropic's public docs and was not run by us. |
| owner-verified | Checked by the maintainer of RoutineOS. It is not a documented pilot. |
| unverified | Not checked. Do not rely on it. Check Anthropic's docs for the current answer. |
Read, then build one
Install the plugin in Claude Code and start from a template. RoutineOS is free (MIT) and needs a paid Claude plan.