How to schedule AI agent tasks in Orchly
Run agents from chat, on a daily, weekly or monthly schedule, or row by row in a table, and choose between Review and Autopilot.
You can schedule AI agent tasks in Orchly, or run an agent three other ways: ask it in chat, run it on rows in a table, or message it in Slack. Scheduled agents run daily, weekly or monthly at a set hour (UTC), and every schedule in your workspace is listed on the Scheduled page.
Separately from when it runs, the agent’s Autonomy decides whether it waits for your approval or publishes on its own.
Run it by hand in chat
Click an agent’s name under Agents in the sidebar and tell it what to do. Or ask Ask Orchly to run it for you: “run the Article Refresher on /pricing”. This is how you test a new agent and how you handle one-off jobs.
Run it on rows in a table
Add the agent as a column in a table and it runs once per row, reading its inputs from your other columns. Press Run on a single row, or run the whole column from its header. This is the best trigger for batch work, like refreshing 30 pages.
How to schedule an AI agent task
There are three places to set a schedule.
1. On the Scheduled page
Switch the sidebar to Agent and open Scheduled. It lists “Everything that runs on a timer, without being asked”. Click New schedule.
- Agent: pick which agent runs. It starts on Ask Orchly.
- What it does each time (optional): the job, in plain language. For example: “Find the 5 pages that lost the most clicks this week and refresh the top two. Report what changed.” Leave it blank and the agent runs its own instructions.
- When (UTC): choose Daily, Weekly or Monthly, then the hour.
- Click Create schedule.
2. On the agent itself
Open an agent’s settings (the gear next to its name) and click Add schedule in the Schedule row. Prebuilt agents allow this too, under their Settings block. The note reads “The agent runs automatically on this schedule.”
3. By asking
Tell Ask Orchly “every Monday at 9, check my AI visibility and post a summary to #marketing”. It creates the schedule and asks you to approve it, since scheduled runs spend credits.
Schedule options
| Cadence | What you set |
|---|---|
| Daily | Every N days, and the hour |
| Weekly | Runs on: one or more days (Su to Sa), and the hour |
| Monthly | On day of the month, and the hour |
Times are on the hour in UTC. If your team is in New York, 13:00 UTC is 9am in summer and 8am in winter.
Manage schedules on the Scheduled page
The table shows every schedule in the workspace:
- What runs: tagged Instruction, Agent or Table column, with the agent’s name underneath.
- Schedule: for example “Daily · 09:00 UTC”.
- Last run and Next run.
- Active: a switch to pause or resume. A paused schedule shows “Paused” as its next run.
Use the delete button on a row to remove a schedule. Past runs stay in the agent’s conversations.
Table column schedules
An agent column in a table can have its own schedule. In the column’s settings, tick On a schedule, set the cadence, and choose Work through N rows each time (up to 50). Each run picks rows the column hasn’t finished yet, so it works down the table and stops when there’s nothing left.
These schedules appear on the Scheduled page with an Open table link instead of a switch. You pause or change them in the table.
A daily schedule that works through 5 rows a day spreads the credit spend and gives you a steady stream to review, instead of 100 drafts on a Monday morning.
Chat with agents in Slack
Once Slack is connected, mention Orchly in a channel or send it a DM. It answers in the thread with the same tools and credits as the app, and asks for approvals there with Approve and Deny buttons.
Review or Autopilot
When nobody is watching, the agent’s Autonomy decides what happens at the end:
- Review: “Produces the work, then pauses for your approval before publishing.”
- Autopilot: “Publishes automatically (per your publishing settings) without asking.”
Most prebuilt agents ship on Review. The Internal & External Link Enricher ships on Autopilot, so check it before you schedule it. See build your first agent for the Autonomy setting.
Run a new scheduled agent on Review for a few cycles. Once its output is consistently good, switch it to Autopilot. A small mistake in a brief becomes a dozen published mistakes on a schedule.
Start runs from your own tools
Scripts and tools like Zapier can start agent work and approve runs waiting in review through the Orchly API. See API keys and MCP.
Before you schedule anything, make sure the agent has the right abilities and a publishing destination.