Skip to main content
Fleetshift provides several safeguards to build confidence before and after code changes are deployed. First, every shift generates diffs that you can inspect before opening any pull requests, so you can review the exact changes the agent proposes. Second, Fleetshift integrates with Soundcheck, which lets you define health checks and compliance standards — you can verify which components pass or fail a given check before and after a shift runs. Third, Fleetshift tracks main-build status after PRs are merged, alerting you if a main branch build fails as a result of the change. You can also target specific subsets of your fleet to validate changes incrementally before rolling them out more broadly.
Portal’s Insights plugin provides a metrics-tracking layer that you can use alongside Fleetshift. While direct Fleetshift-to-Insights integration is on our roadmap, you can already use Insights to monitor engineering and operational metrics across your fleet. Fleetshift itself tracks leading indicators such as PR merge status and main build pass/fail through its shift dashboard. For broader business and cost metrics, you can configure Insights to surface the signals that matter to your organization and track them over time as shifts progress.
Fleetshift works alongside your existing SCM and CI/CD infrastructure. We currently support GitHub and Azure DevOps as SCM providers, with GitLab coming soon. For AI providers, Fleetshift supports Anthropic (Claude), Amazon Bedrock, Google Vertex AI, and OpenAI. Fleetshift opens pull requests that flow through your standard review, CI, and deployment pipelines — this means all existing tests, checks, and monitoring continue to run as they would for any other code change. Main build status is tracked automatically via webhooks after PRs are merged, so you can detect regressions early regardless of your underlying infrastructure.
For security vulnerabilities at scale, Fleetshift offers both agentic and deterministic approaches. In agentic mode, the AI agent (Honk) analyzes each repository’s code individually and generates tailored fixes. In deterministic mode, you can use OpenRewrite recipes or npm package shifts for well-defined, repeatable transformations. To validate fixes before a fleet-wide rollout, you can test your shift against a small subset of repositories first — review the generated diffs, verify that CI builds pass, and confirm that Soundcheck checks transition from fail to pass.
Yes. Fleetshift supports targeting any subset of your catalog entities — you can filter by entity type, tags, ownership, or Soundcheck check status. We recommend a tiered rollout strategy: start with lower-risk services, verify that PRs merge cleanly and main builds pass, then proceed to the next tier. You can also use Soundcheck campaign insights to monitor your fleet-wide progress in real time as PRs are merged and checks transition from fail to pass.
Fleetshift supports four types of shifts: Agent shifts use an AI coding agent (Honk) to implement changes based on a natural-language prompt — ideal for complex migrations like converting JavaScript to TypeScript or fixing nuanced security issues. Dependency update shifts handle version bumps for npm, Maven, pip, Go, and other package managers. OpenRewrite shifts run community or custom OpenRewrite recipes for structured refactoring like API renames, framework migrations, and dependency upgrades. npm package shifts let you publish deterministic transformation scripts as npm packages for fully repeatable, token-free changes.
Fleetshift and Soundcheck work closely together. From a failing Soundcheck check or campaign, you can click “Fix with Fleetshift” to launch a shift with the failing components pre-filled as targets. After PRs are merged, the check status updates automatically from fail to pass. For campaign-level fixes, you can track all created PRs and watch the campaign pass rate increase as PRs land. You can also use “Fix with AiKA” to conversationally diagnose and remediate a failing check on a single component before scaling the fix fleet-wide via Fleetshift.