Automations

Automations trigger cloud agents automatically based on events in your repository or connected services. Turn repetitive workflows into one-time setups.

Cloud Agents required

Automations run via Cloud Agents. A Vyre Pro or Enterprise subscription is required.

Trigger types

Creating automations

Go to Settings → Cloud Agents → Automations and click "New Automation."

yaml
# .vyre/automations/review-pr.yaml
name: Auto-review PRs
trigger:
  event: pull_request.opened
  filter:
    base_branch: main
    labels:
      exclude: ["skip-review", "wip"]

agent:
  task: |
    Review this pull request thoroughly.
    1. Check for logic errors and edge cases
    2. Verify test coverage for new code
    3. Check for any security issues
    4. Write inline review comments
    5. Post a summary review with overall verdict
  model: claude-sonnet-4-5
  max_duration: 300  # 5 minutes

Automation examples

@vyre mention handler

The most powerful automation. Any team member can ask Agent to do something by mentioning @vyre-agent in a comment:

yaml
name: Handle @vyre mentions
trigger:
  event: issue_comment.created
  filter:
    body_contains: "@vyre-agent"

agent:
  task: |
    Read the comment that mentioned you. Understand what
    the commenter wants and take the appropriate action.
    Reply to the comment with what you did.

Example usage in a PR comment:

"@vyre-agent please add tests for the new endpoint and update the README"

Nightly dependency updates
yaml
name: Weekly dependency updates
trigger:
  schedule: "0 9 * * 1"  # Every Monday at 9am UTC

agent:
  task: |
    Run `npm outdated` to find outdated packages.
    Update packages that have only patch or minor version bumps.
    Run the test suite to verify nothing broke.
    Open a PR with the changes titled "chore: update dependencies [date]"
CI failure investigator
yaml
name: Investigate CI failures
trigger:
  event: check_run.failed
  filter:
    workflow_name: "CI"
    branch: main

agent:
  task: |
    A CI run on main just failed. Read the failure logs,
    identify the root cause, and open a high-priority issue
    with a detailed description of what broke and why.
    Label the issue "ci-failure" and assign it to the
    last committer to main.