Full Review and JSON Output

Codity CLI can run a combined local review and return either human-readable terminal output or machine-readable JSON.

Use this when you want one command to cover review, security, and code quality, or when an automation tool needs parseable results.

Full Review

codity review --full runs three checks for the same diff:

  • AI code review for bugs, regressions, and maintainability issues.
  • Changed-code security checks and vulnerabilities introduced by dependency changes, including supported transitive lockfile resolutions.
  • Code quality scanning for complexity, maintainability, and anti-patterns.
codity review --full --staged

You can combine --full with the same diff-selection flags as normal review:

codity review --full --all
codity review --full --commit abc1234
codity review --full --branch main

JSON Output

Use --json when another tool needs to parse the result:

codity review --json --staged

For one combined machine-readable result:

codity review --full --json --staged

The JSON output includes:

  • status - Review status, such as completed or no_changes.
  • summary - Human-readable summary of the review.
  • pr_summary - Pull-request style summary when available.
  • comments - Actionable review findings.
  • counts - Findings in comments, by severity. Does not include security or quality findings, so it can read 0 while those hold critical issues.
  • security - Security findings when --full is used.
  • quality - Code quality findings when --full is used.

Example Workflow

git add .
codity review --full --json --staged

Use the JSON output in CI, local scripts, or agent workflows that need structured findings without parsing terminal formatting.

Notes

  • If no changes are found, JSON output returns status: "no_changes" with an empty comments list.
  • Approval-only comments are filtered out so the result focuses on actionable findings.
  • In normal terminal mode, Codity still streams progress and renders findings in a readable format.

A successful command exit is not a clean-review verdict. Inspect comments, security.findings and quality.findings together. The counts field summarizes review comments alone. Some CLI commands accept a global --json flag without implementing JSON output; for combined scan automation, use this documented review flow.

Repository audits and dependency coverage

Full repository scans and license reports start only in dashboard Security. review --full checks the selected change and never falls back to a repository audit. Existing repository dependency debt is excluded.

Dependency analysis compares complete before/after evidence. Attribution supports public npm lockfiles versions 2 and 3, public PyPI Poetry/Pipfile locks, crates.io Cargo locks and RubyGems locks. Exact-pinned requirements can report direct introductions with partial transitive coverage. Unsupported formats, missing base evidence or unresolved versions show partial or unavailable coverage. Advisory failures show failed coverage. No dependency changes are not applicable. Those states do not promise a clean repository scan.

security.dependency_analysis exposes the reviewed base/head/workflow, coverage status and reasons. Each dependency finding includes matching change_evidence with the before/after resolution, advisory aliases and dependency paths. Older clients without complete file pairs receive unavailable dependency coverage. Inspect that status even when security.findings is empty.