Skip to main content
These guides walk through common Fleetshift workflows end to end. Each one covers a specific use case and can serve as a starting point for your own shifts.

​
Create and run an agent shift

Use this guide to walk through the core shift lifecycle — creating a shift, running it against target repositories, and merging the resulting pull requests.
  1. Navigate to Fleetshift in the sidebar.
  2. Click New Shift and select Agent.
  3. Fill in:
    • Name: add-codeowners
    • Prompt: Add a CODEOWNERS file to the root of the repository based on the existing contributors.
    • PR title: Add CODEOWNERS file
  4. Select two or three target repositories and click Create.
  5. On the shift details page, select all targets and click Run.
  6. Watch the target status transition from pending to jobRunning to jobComplete.
  7. Open the Logs tab to follow the agent’s progress in real time.
  8. Once complete, open the Diff tab to review the generated changes.
  9. Click Create PR on a completed target.
  10. The status transitions to prCreating and then prOpen. Click the PR link to open it in GitHub.
  11. After you merge the PR in GitHub, the status automatically updates to prMerged through the webhook you configured during setup.

​
Fix a failing Soundcheck check

If a component is failing a Soundcheck check, you can launch a Fleetshift shift directly from the Soundcheck UI to generate a fix.
  1. Navigate to Soundcheck and open a check that has failing components.
  2. On the Check Insights page, find a failing component.
  3. Click Fix with Fleetshift.
  4. You are taken to the Fleetshift new shift wizard with the Soundcheck check pre-filled as the target source.
  5. Write a prompt that describes the fix, such as: Fix the failing check by adding the required configuration file.
  6. Click Create, then trigger the shift and wait for it to complete.
  7. Create a PR from the completed target.
  8. Back in Soundcheck, the check details page shows the Fleetshift PR with its current status. Click View Progress to jump to the shift details.
  9. After the PR is merged, the Soundcheck check transitions from Fail to Pass and the Fix with Fleetshift button is hidden.

​
Fix a failing check with AiKA

You can also remediate failing Soundcheck checks conversationally through AiKA.
  1. Navigate to Soundcheck and open a check with failing components.
  2. On a failing component, click Fix with AiKA.
  3. The AiKA chat panel opens with the check context preloaded — including the check name, entity ref, and failure details.
  4. AiKA analyzes the failure and proposes a fix. Follow its guidance to remediate the issue.
Fix with AiKA is only available when AiKA is configured on your Portal instance.

​
Run a campaign-level bulk fix

When a Soundcheck campaign has many failing components, you can use Fleetshift to fix them all at once instead of one at a time.
  1. Navigate to Soundcheck and open a campaign with multiple failing components.
  2. On the Campaign Insights page, click Fix with Fleetshift.
  3. The Fleetshift wizard opens with the campaign as the target source, showing all failing components.
  4. Create and trigger the shift. Jobs run against all failing components in parallel.
  5. Back on the Campaign Insights page, a Shift in progress label indicates the shift is running.
  6. The Pull Requests table shows all created PRs with their status, component name, and links. You can sort and filter by component name or status.
  7. As you merge PRs in GitHub, the campaign pass rate increases automatically as checks transition from Fail to Pass.

​
Use MCP actions for an AI-assisted workflow

If you use AiKA or another MCP-compatible AI agent, you can drive the entire shift workflow through conversation. See the MCP actions reference for the full list of available actions. Here is an example conversation flow:
  1. Ask your agent: What Soundcheck checks are failing on <component name>? The agent calls find-shift-for-soundcheck and returns matching shifts or suggests creating one.
  2. Ask: Fix the failing <check name> check on <component name>. The agent calls trigger-honk-planning and presents a remediation plan for your approval.
  3. After you approve, the agent calls trigger-honk-job and polls get-job-status to report progress.
  4. When the job completes, the agent calls create-prs and shares the PR link.
  5. At any point, ask How is my shift doing? and the agent calls get-shift-stats to report completion percentage, merge rate, and failure rate.