Scheduled Agents

A lot of security work is the same job on a loop: scan the repo, check open findings, tighten guardrails. It keeps coming back, on a cadence and when new findings land. Scheduled Agents run those jobs for you. They use Corridor Agent.

​

Where to find them

Open Agent → Scheduled Agents. You can:

• Create an agent from a template or a custom prompt
• Pick a time schedule or a findings trigger
• Choose one project or many
• Run an agent now
• Pause or resume it
• Open run history

​

Templates

​

When an agent changes a guardrail

Improve Guardrail Coverage and Threat Model write guardrails. A custom agent can do the same if you ask it to. Once an agent can edit a guardrail, you need to see what it changed and go back if the edit is wrong. That is guardrail versioning. Each guardrail keeps an append-only history. In the app, open the vN chip. Through MCP, Owner and Admin can list versions with getGuardrailVersions and restore one with restoreGuardrailVersion. An AI assistant can also change guardrail text and archive a guardrail through MCP. Use updateGuardrail to edit name, overview, or text. Use archiveGuardrail to stop a guardrail from running. That write path is a separate private beta. See Update and archive through MCP. Un-archive stays on the dashboard. Restoring a version does not un-archive a guardrail.

​

When an agent runs

​

Time schedules

Pick daily, weekly, or monthly. Times are Pacific Time. Use a time schedule for work you want on a cadence, such as a weekly full scan.

​

Findings triggers

You can also run an agent when findings change:

• Finding created
• Finding status changed

These triggers only fire for findings on the main branch. They only fire for finding events after you activate the agent. Use a findings trigger when you want work to follow new issues, such as tightening guardrails after a scan. An agent uses a time schedule or a findings trigger. It cannot use both.

​

Project scope

Pick one project or several. Leave the list empty to cover every project on the team.

​

Run history

Each run shows when it started, its status, and a link to the chat when one exists. Status is Dispatched, Missed, Skipped, or Failed. Open the chat to read what the agent did.

​

Next steps