Vet

About Vet

Vet is a fast, local code review tool released as open source that checks work produced by AI coding agents. It verifies agent outputs against the conversation history to ensure actions match requests and flags silent failures like half-implemented features or tests that were never actually run.

Review

Vet emphasizes concise, targeted feedback rather than long, noisy reports. It can run from the CLI, be integrated into CI pipelines, or be used as an agent skill, and it supports local models with zero telemetry for privacy-conscious workflows.

Key Features

• Agent-aware reviews: inspects conversation history to confirm that an agent's actions align with the requested goals.
• Silent-failure detection: catches issues such as fake test runs, partial implementations, and claimed but unexecuted work.
• PR-level analysis: reviews full diffs for logic errors, unhandled edge cases, and deviations from stated objectives.
• Local-first operation: works with existing API keys and local models, and reports no telemetry by default.
• Flexible deployment: run from the command line, in CI, or as a skill within an agentic workflow.

Pricing and Value

Vet is free and open source. Its value comes from automating parts of the verification process for teams that rely on AI coding agents, reducing the manual effort required to spot subtle failures. Performance and cost scale with diff size and the model/context used; reported times are low in typical configurations (on the order of seconds for many diffs), though larger diffs may increase runtime and resource use.

Pros

• Open source and auditable, so teams can inspect or extend checks to fit their codebase.
• Operates locally with no telemetry, which helps with privacy and sensitive projects.
• Targets a specific gap in AI-assisted development by verifying agents against conversation history.
• Concise output aims to surface relevant issues without excessive noise.
• Integrates into existing workflows via CLI and CI support.

Cons

• Performance and cost increase with very large diffs or when hitting model context limits.
• Effectiveness depends on the quality and completeness of the agent's conversation history and the model used.
• May require configuration or added checks to match project-specific standards and coding conventions.

Vet is best suited for developers and teams that already use AI coding agents and want an automated layer of verification integrated into local and CI workflows. It is a practical tool for catching routine alignment and execution errors, though complex architectural reviews still benefit from human oversight.

Open 'Vet' Website