Pre-merge database linter + enforcer + safety net.
Detects risky patterns, stops dangerous migrations, and gives helpful feedback before you merge code.
Valk Guard analyzes your codebase using static analysis, meaning it doesn’t need to run your app or connect to a live DB.
Uses an internal SQL parser (ANTLR-based) to detect destructive or locking changes.
Traverses source code using language-specific ASTs to catch risky patterns.
Each issue it detects is checked against a rule set defined in .valk.yaml. No AI guesswork.
Fails the CI job instantly. Used for destructive changes or critical performance risks.
Posts a comment on the PR but lets the build pass. Good for style checks or minor risks.
Skips the check entirely for specific files or lines. Configurable in YAML.
Run Valk Guard as a CLI tool locally, a GitHub Action, or a step in GitLab CI / CircleCI.
Valk scans changed files in PRs, producing inline comments or blocking failures.
When Valk finds something ambiguous — like a complex query — it provides a "Run Simulation" link.
Send the query to Valk’s cloud service to get deterministic cost estimation and understand why the issue matters.
By default, Valk Guard runs entirely locally or in your CI, without sending any code, schema, or data outside your environment. Simulation is opt-in only.