Summary

Vet is a standalone verification tool for code changes and coding agent behavior, ensuring goal adherence and code correctness. It runs anywhere—terminal, agent skill, or CI—and supports any LLM provider via user API keys, offering a free, open-source solution without data collection.

How It Works

Vet analyzes code changes and agent conversations by snapshotting the repository, diff, and optionally agent interactions. It then leverages user-configured LLMs to perform checks, filtering findings into a deduplicated list of issues. This approach provides flexible, provider-agnostic verification directly through the user's inference endpoints.

Quick Start & Requirements

Install the CLI via pip install verify-everything, pipx install verify-everything, or uv tool install verify-everything. An agent skill can be installed via a provided script (curl -fsSL https://raw.githubusercontent.com/imbue-ai/vet/main/install-skill.sh | bash), supporting project or user-level installations. Vet integrates with GitHub Actions using a reusable workflow, requiring repository secrets for LLM API keys (e.g., ANTHROPIC_API_KEY). Users must provide their own API keys for supported LLM providers.

Highlighted Details

• Agent Integration: Agents automatically run Vet post-code changes for proactive issue detection.
• Bring-Your-Own-Model (BYOM): Works with any LLM provider via API keys, including OpenAI and Anthropic, configurable through JSON files and a remote model registry.
• CI/CD Ready: Features a reusable GitHub Action for automated pull request reviews.
• Extensible Configuration: Supports custom model definitions, configuration profiles (TOML), and customizable issue guides (TOML) for standardized team usage.

Maintenance & Community

Developed by Imbue, Vet encourages community engagement through its Discord server for discussion and support, and GitHub Issues for bug reports and feature requests.

Licensing & Compatibility

Licensed under the GNU Affero General Public License v3.0 (AGPL-3.0-only). This strong copyleft license requires derivative works to be shared under the same terms, potentially impacting integration into closed-source commercial products.

Limitations & Caveats

The --history-loader option executes shell commands, necessitating careful review of associated commands and configurations. Users are responsible for managing their LLM API keys and associated inference costs. The AGPL-3.0 license imposes significant obligations on derived works.