Open WebUI added scheduled runs for coding agents in its November 8, 2026 release, giving software maintainers and small development teams a way to repeat tasks without launching each run manually. Version 0.13.3 extends the application’s existing automations; scheduling prompts itself is not new.
Users can save a terminal and working folder, then choose a coding agent available there. The terminal is the computing environment containing the files and tools the agent uses. One possible application is a daily summary of recent project changes for a maintainer to review.
What needs to stay available
At execution time, the scheduler checks that the saved terminal and agent are still available. Agent results must go to a chat, not a channel. The background worker must be running, and the owner must still have permission to use automations.
For a team using a development machine, the practical question is whether that environment will be available at the scheduled time. Saving an automation is not a substitute for arranging the machine, agent access and usable project folder.
Open WebUI’s license guidance describes the core software as free to use, subject to its conditions, including branding rules. That does not establish free model usage or hosting; the release gives no all-in price per run.
Choose a task that can stand alone
A saved working folder is not saved conversation history. The scheduler creates a new chat for each run, while agent-session resumption depends on prior conversation state. The practical implication is to put the recurring objective and necessary context in the saved instructions, rather than relying on yesterday’s chat.
A timetable also does not guarantee completion without a person. The agent approval code can leave requests pending, and questions to the user need an answer even when tools have full access. Check the automation’s actual permission settings and decide who will respond before relying on background work.
A bounded reporting task is a reasonable starting point. Use Run Now and execution history to evaluate your setup before relying on a recurring schedule. A success entry records that the reply finished—not that any suggested code is correct or approved.
Keep the decision to accept changes separate from the scheduled run. RohitAI’s coverage of Copilot’s review canvas in Teams explores that distinction in a different coding-agent workflow.