AI Review CLI
A CLI that reviews pull requests in CI: diff comments, a summary and thread replies via OpenAI, Claude, Gemini or Ollama.
- 607 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The diff and repository context are sent to the chosen model provider.
- The VCS token lets the tool post and delete comments in pull requests.
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/ai-reviewDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The diff and repository context are sent to the chosen model provider.
- The VCS token lets the tool post and delete comments in pull requests.
This is third-party code. Review the repository files before installing.
What it does
AI Review runs in CI on a pull or merge request, reads the changes and leaves a language-model review in them. `ai-review run-inline` writes line-by-line comments in the diff, `run-context` looks for problems across several files, `run-summary` posts a short wrap-up, and `run-inline-reply` and `run-summary-reply` answer in discussions already under way. Each mode has a `clear` command that removes what it left. The model is chosen from OpenAI, Claude, Gemini, Ollama, Bedrock, OpenRouter and Azure OpenAI and switched in settings. The version-control system plugs in from GitHub, GitLab, Bitbucket Cloud, Bitbucket Server, Azure DevOps and Gitea. Configuration lives in YAML, JSON or environment variables, and prompts for inline, context and summary reviews can be rewritten to match team rules. In agent mode the model, in a ReAct-style loop, explores the repository itself with `ls`, `cat`, `rg` and `git` before giving its verdict, getting more context than a single call. The author stresses that the tool runs entirely client-side: requests to the model and VCS go directly, with no intermediate proxy.
Who it is for. Teams on GitHub, GitLab or Bitbucket that want automatic AI review of pull requests with their own key and chosen model rather than a cloud service.
Good fit when
- Reviewers spend time on small remarks and a first model pass over the diff is wanted.
- A general verdict on a large pull request with a list of main risks is needed.
- An author wants a model reply right in the discussion thread under a reviewer's remark.
- Review has to run in GitLab CI from a ready Docker image.
- A team must switch the model from OpenAI to local Ollama without reworking the pipeline.
Not a fit when
- Code must not go to an external model provider: with cloud models the diff is sent to the provider.
- You need strict deterministic linting: model review varies between runs.
- Review is meant to replace a human decision: the tool leaves comments and merging stays with people.
Example request
Add a manual AI Review run to our GitHub Actions for a pull request with the gpt-4o-mini model.Limitations
The result depends on the chosen model, and remarks can be inaccurate or false, so a human reads the comments.
Context review needs the full Git history: set fetch-depth 0 or GIT_DEPTH 0 in CI.
Model keys and the VCS token pass through environment variables or CI secrets, and in agent mode the model runs `ls`, `cat`, `rg` and `git` in a copy of the repository.
The author takes support on Boosty, and the README has a section on privacy and responsibility for the code that is sent.
How to disable. Remove the ai-review job from the workflow or `.gitlab-ci.yml`. Delete the comments it left with `ai-review clear`, `clear-inline` and `clear-summary`.
FAQ
How do I install AI Review?
With `pip install xai-review` or by running the Docker image `nikitafilonov/ai-review:latest` with `ai-review run-summary`. In GitHub Actions it is used as `Nikita-Filonov/ai-review@v1.4.0`.
How do the review modes differ?
Inline comments on lines, context finds problems across files, summary gives a short wrap-up, and reply modes continue discussions under a reviewer's comments.
Where does the code go?
Directly to the chosen model provider and your VCS. The author states that the tool runs client-side and neither proxies nor inspects requests.
Can I use a local model?
Yes. Ollama is supported, as are Bedrock, OpenRouter and Azure OpenAI, and the provider is switched with the `LLM__PROVIDER` variable.
How do I remove the comments it left?
With `ai-review clear` for everything or `clear-inline`, `clear-summary`, `clear-inline-reply` and `clear-summary-reply` for single modes.
Discussion
No questions or reviews yet.
Related
CLI toolsOpen-source AI agent for pull request review: descriptions, feedback and improvement suggestions on GitHub, GitLab and more
Reviews the git diff and the conversation with the agent: catches bugs and gaps between your request and the agent's result. CLI, skill, GitHub Action
Snapshot tests for AI agents: records tool calls today and tells you when they silently change after an edit
A check for AI-generated code: empty functions, imports of things that do not exist, and docs that oversell.
An ESLint-style linter whose rules are judged by a model: it finds in agent-written code what syntax checks miss and sets a verdict for CI
Runs Claude Code, Codex, Gemini and OpenCode on YAML tasks in a sandbox, scores the result and gates CI on it
README in short
The README has a table of contents, a support section, a feature description, live pull request examples for each mode, quick start, GitHub Actions and GitLab CI templates, documentation links and a privacy note.