Product Overview

Burg AI is an AI-powered code review assistant that automatically analyzes GitHub pull requests(for now). It runs when PRs are opened or updated, posts review comments directly on GitHub, and helps engineering teams catch issues before merge.

Problem Solved

Human code review is slow, inconsistent, and often misses edge cases. We provide:

• Immediate feedback — Reviews get completed in seconds, not hours.
• Consistent analysis — Same standards applied to every PR.
• Deeper inspection — Checks patterns humans often overlook, and remembers them.

What Happens Inside a PR

When you open or update a pull request:

1. Burg AI receives a webhook from GitHub.
2. It fetches the PR diff and file contents.
3. The code is analyzed using your configured LLM (via BYOK) and the repo-level settings.
4. A summary about what the PR is about, with a few comments if required for any issues that the agent discovers in its analysis.

What Burg AI Does NOT Do

• Does not replace human reviewers — It supplements engineering teams and preserves final decision-making with humans.
• Does not block merges — Merges proceed normally unless explicit approval gates are configured.
• Does not modify code — It only posts review comments on pull requests and never pushes commits.
• Does not train on your code — Code is processed transiently using your configured LLM providers and discarded immediately after PR analysis.
• Does not store source code — Only minimal, non-code metadata required for system operation is retained.

How It Works (High-Level)

┌─────────────┐    ┌─────────────┐    ┌─────────────┐
│  GitHub     │───▶│  Burg AI    │───▶│  LLM        │
│  Webhook    │    │  Server     │    │  (Your Key) │
└─────────────┘    └─────────────┘    └─────────────┘
                          │
                          ▼
                   ┌─────────────┐
                   │  GitHub API │
                   │  (Comments) │
                   └─────────────┘

1. GitHub App installed — Burg AI is installed as a GitHub App on the selected repositories
2. PR opened or updated — GitHub triggers a webhook event
3. Agent runs — The diff is analyzed using your BYOK LLM configuration
4. Review posted — Inline review comments are added directly to the pull request
5. Feedback loop — Developer reactions are used to calibrate future review behavior

Next Steps

• Quick Start — Install and run your first review
• Core Concepts — Understand the terminology