Straight answers about cost, safety and Claude routines.

It is Claude Code work you set up once and Anthropic's cloud runs for you on a timetable, even when your laptop is closed. RoutineOS adds a Desk where you decide, approval gates, typed settings and a loop that suggests improvements.

Yes, two names for one feature. Anthropic's docs now say scheduled tasks, formerly routines. In RoutineOS, the routine is the folder, Desk and rules you build, and the scheduled task is where it runs.

Install the RoutineOS plugin in Claude Code, ask for the job in plain words, then schedule it. A daily schedule is within the limit. The shortest interval is one hour, in UTC, and a run can start a few minutes late.

No. A cloud run does not load plugins installed on your machine. Only skills committed to the repository load there. RoutineOS commits what a routine needs to its own folder so the scheduled run can use it.

Not as a like-for-like swap. n8n runs fixed flows. A RoutineOS routine is a Claude job with a Desk and approval gates. If you have an n8n workflow, RoutineOS can convert it and lists anything it could not carry over for you to finish.

No. Describe what you want in plain words, Claude sets it up, and you check the first run. Claude Code is the tested route. A Claude Desktop install exists but is untested for scheduling.

The prompt holds anything that sends, posts or deletes until you say yes. The first run is report-only. Text a routine reads is treated as information, not orders. It never applies its own suggestions and never asks you to paste a secret into the chat.

Not by design. The prompt holds sending, posting and deleting at an approval gate until you approve. On a report-only first run, the prompt forbids sending at all.

It tracks what you keep and what you drop. When your choices point the same way, it proposes one small change and shows its reasoning. You answer yes, no or later.

Free (MIT). Needs a paid Claude plan. RoutineOS is open source, and your routines use that plan's usage. The daily run cap is a rule in the prompt, not a hard guarantee.

Once an hour at most, in UTC. A faster schedule is rejected. There is no Run now button, so a request made on the Desk waits for the next scheduled run. Scheduled tasks are a research preview, so limits can change.