Run a local AI code review trial before you add a team CI gate

AI-assisted code review is easy to demo. The harder question for a team is whether a repeatable rule set will catch something useful before merge — without turning every pull request into an opaque SaaS workflow.

I built a small, local-first trial for that question. It takes about 10 minutes and uses a deliberately unsafe sample, not your repository.

What the trial proves

Run the included script:

./run-trial.sh

The sample produces a structured local report with:

• a High command-injection finding for os.system(user_input);
• a Medium team-convention finding for a direct print call; and
• exit status 1 for high or medium findings. Critical findings use exit status 2.

That is intentionally narrow: this is deterministic local pattern checking with structured findings, not a claim to cover an entire security framework. The point is to see whether the workflow and rule quality fit your team before you standardize it.

Where the free server fits

The MIT-licensed aicraft-code-review server can review code, diffs, and local files from MCP clients such as Claude Code, Cursor, and Cline. It gives a team a local place to start testing shared review behavior.

The standalone trial uses only its included sample. It creates no payment or subscription.

• Run the free trial: https://aicraft.vip/trial.html
• Full walkthrough: https://aicraft.vip/blog/team-code-review-trial-en.html

When a team needs a rules pack

If the free trial catches something worth enforcing everywhere, the optional Team Rules Pack is a one-time $49 purchase. It includes 63 ready-made rules for Python, JavaScript/TypeScript, Go, and Java; GitHub Actions and GitLab CI merge-gate templates; and 20 LLM review prompts for security, reliability, performance, and test gaps.

It is intended for teams that need a versioned baseline rather than a different prompt on every machine. Secure checkout: https://www.creem.io/payment/prod_6Z3S3jGNPsCyRSqNi397ZY

If you try the free workflow, I would value three specific data points: what it caught, which rule your team would share, and which result was noisy or missing. A structured team-trial form is here: https://github.com/GoodJobwilliam/aicraft/issues/new?template=team-trial.yml&title=Team%20trial%20request

No source code or credentials are needed in feedback.