Visual Studio Code 1.145, released November 4, 2026, lets makers schedule local AI agent automations with a five-field cron expression and choose a custom agent from the current workspace. GitHub called out the change again in its November 6 Copilot weekly roundup. A scheduled agent can revisit a repository task without the maker starting each run manually, but the schedule is only a trigger. The task still needs a clear prompt, an appropriate agent and a reviewable result. VIBPS has not run a VS Code 1.145 automation firsthand.
What is new in the automation schedule?
VS Code's release notes say the local automation form now accepts a cron schedule when its presets are too coarse. The editor validates the expression and describes when it will run. Makers can also choose a workspace custom agent when creating or editing a local automation. Disabled automations move into a collapsed section to keep the active list readable. The cron option is for local automations; Microsoft's documentation says it is not available for cloud automations. The release does not make every local job run while the computer is off.
Local or cloud: what has to stay running?
Microsoft's automation documentation says local runs need the development environment: the computer must stay awake and the agent must be available at the scheduled time. Cloud automations run on GitHub even when the maker's computer is off, but require an eligible repository, service access and the relevant account permissions. Cloud automations remain in preview and have their own schedule and event options. Choosing local execution is sensible when a task depends on files or tools on that machine; choosing cloud execution changes the environment and access model. A maker should not assume a local cron expression turns a laptop into an always-on service.
The setup flow starts in the Agents window's Automations view. A maker provides a name and prompt, selects the workspace and agent, chooses permissions and a trigger, then reviews the configuration before creating it. Microsoft suggests worktree isolation for Git tasks when the agent supports it. For an app project, a useful scheduled job might summarize new issues or check whether documentation needs updating. Giving an unattended task broad production access would be a separate decision, not a property of the cron feature.
What else arrived for agent-built apps?
VS Code 1.145 also previews browsing an app through a remote agent host from the Agents window. With the relevant setting enabled, a maker can open a development server running on the remote host and inspect the app without leaving the editor. Remote canvases are also in preview behind their settings. That can make an agent's change easier to check, but the release notes explicitly label these as previews. A remote preview is a validation surface, not proof that a production app works across devices or for real users.
The broader point for makers is that recurring agents can maintain a software project while humans set its scope and inspect its output. VIBPS is the social network for software, where an app can be shared with people who may follow its creator after it is ready to use. A schedule can help the build process; a working product and a clear account of what changed remain the maker's responsibility.