Nayjest/Gito

Gito – AI‑powered code reviewer

What it is – Gito is an open‑source tool that uses any LLM (OpenAI, Anthropic, Google, local models, etc.) to automatically review code changes. It can run as a CLI on a developer’s machine or be triggered from CI/CD pipelines (GitHub Actions, GitLab CI) to comment on pull‑requests.

Key capabilities

• Vendor‑agnostic LLM access – works with any provider or self‑hosted model via the ai‑microcore abstraction.
• Fast, parallel review – sends files to the model concurrently and returns a markdown report in seconds.
• Broad language support – claims to handle “all major programming languages and frameworks”.
• Configurable – per‑machine .gito/.env for credentials and per‑project .gito/config.toml for prompts, severity levels, excluded files, etc.
• Integrations – built‑in GitHub and GitLab CI workflows, Jira & Linear issue‑tracker hooks, and a placeholder for Bitbucket (planned).
• Privacy‑first – stateless client‑side tool; code is sent only to the configured LLM endpoint, never stored by Gito.

Getting started

1. Install – pip install gito.bot (or use uvx gito.bot or the Windows installer).
2. Configure – run gito setup to create ~/.gito/.env with your LLM API key, model name, etc.
3. Run locally – gito review from the repository root to get a markdown report.
4. Add to CI – add the provided GitHub Actions workflow (.github/workflows/gito-code-review.yml). The workflow installs the package, sets the LLM secrets, runs gito review, and posts the results as a PR comment.

Typical workflow

• A PR is opened → GitHub Action triggers → Gito sends the diff to the LLM → LLM returns findings (security, performance, style, etc.) → Goto posts a comment with a formatted report.

Known limitation

• Gito cannot modify workflow files from within a PR comment when using the default GITHUB_TOKEN (GitHub blocks such changes). A personal access token with workflow scope would work but is discouraged for security reasons.

Where to learn more

• Full command‑line reference, configuration cookbook, and platform‑specific setup guides are in the documentation/ folder of the repo.

License – MIT.