Built by Vineet Kumar · v0.1

CodeReview Agent

AI-powered code review for Quality Engineering teams. Paste a GitHub or GitLab URL — get AI-flagged style and convention violations in seconds. If your GitLab URL needs authentication, use a public GitHub repository URL (or an NPM package repository link), or download the source code and review it locally.

☕ Server is waking up from sleep — first load may take 30–60 seconds. Please wait… ☕ Welcome to CodeReview Agent. The server is waking up from sleep — first load may take 30–60 seconds.
Don't have one handy? Try a sample PR ↗
wakeupBanner.style.display = "flex";
Supported URL formats https://github.com/<owner>/<repo>/pull/<number>
wakeupBannerMessage.innerHTML = "CodeReview Agent is ready. Your first review is about to start."; wakeupBannerIcon.textContent = "✓"; GitLab: https://gitlab.com/<owner>/<repo>/-/merge_requests/<iid>
        https://gitlab.com/<owner>/<repo>/-/commit/<sha>
wakeupBanner.style.display = "flex"; (Public repos only in this phase — private repo support coming }, 2200); >
Quick walkthrough

See it in action

A 15-second walkthrough — paste a pull request URL, the agent fetches the diff, runs it through the LLM against your standards, and returns structured findings.

codereview-agent.app DEMO
Pull request URL
Review →
Fetching diff from github.com…
17 issue(s) found (7 high, 10 medium)
github · pr
▸ tests/test_violations_sample.py 17 ISSUES
High line 11
Hardcoded API keys must be replaced with environment variables or a secret manager.
security · secrets
Medium line 24
Use pytest.fixture instead of setUp/tearDown for test setup consistency.
framework · pytest
High line 38
Bare except: swallows real errors — catch a specific exception type.
style · error-handling
Medium line 52
Avoid time.sleep() in tests — use explicit waits or mocks for deterministic runs.
convention · flakiness

What this is

An AI-powered code review agent that demonstrates a pattern used in production at regulated banking environments. Built as an open-source illustration — paste any public GitHub or GitLab URL, and the agent will flag style, convention, and framework standard violations using an LLM, scoped to these Quality Engineering standards.

The agent runs against your team's coding standards (configurable), uses any LLM provider you prefer (Gemini, Ollama, OpenAI, Claude), and integrates with CI as a GitHub Action. This hosted demo uses Gemini's free tier.

Pluggable LLM

Same agent code works with Gemini, Ollama, OpenAI, or Claude. Swap providers via one config line.

Structured output

Returns parseable JSON with file path, line number, severity, category, and message — ready for CI gates.

Standards as data

Coding rules live in a markdown file. Forking and customizing for your team takes 5 minutes.

Honest framing

Style and convention enforcement — not architectural or logic review. Designed to free up senior reviewer time.