Skip to main content
Macroscope automatically reviews every pull request for quality and correctness. When it finds an issue, it adds a GitHub comment describing the bug with a suggested fix. Engineers can give feedback with 👍 or 👎 reactions to improve future reviews. Bug Finding and Fixing

​
Configuration

  • Code review is enabled by default on every linked repo. Toggle per-repo in Settings → Repos.
  • Change the default for newly linked repos in Settings → Repos > Defaults.

​
Manual Triggers

Comment “@macroscope-app review” on any PR to trigger a review, even if code review is disabled for that repo.

​
Severity Levels

Each comment is labeled with a severity:
  • CRITICAL: data loss, security breach
  • HIGH: production crashes, security degradation
  • MEDIUM: broken functionality (recoverable)
  • LOW: cosmetic, edge-case issues

​
Labels

  • Always-review labels: trigger code review even when automatic review is disabled. Configure in Settings → Repos.
  • Skip labels: prevent review and PR summaries, even when automatic review is enabled. Skip labels take precedence over always-review labels. Configure in Settings → Repos.

​
Comment Resolution Tracking

Comments link to the commit where they were resolved. Macroscope automatically re-evaluates when new code is pushed.

​
Check Run Status

When issues are found, the GitHub check run completes with NEUTRAL instead of SUCCESS. Check run details show per-file exclusion reasons (file type, size, skip patterns).

​
AI Web Research

Macroscope searches the web for latest documentation on libraries you use, ensuring reviews account for up-to-date API signatures, deprecation notices, and best practices.

​
How Code Review Works

Macroscope uses code walkers that traverse the Abstract Syntax Tree (AST) to build a graph-based representation of your codebase. This gives Macroscope deep understanding of how each change impacts your product, enabling it to catch bugs that tools with shallower analysis miss.

​
Supported Languages

Native AST-based analysis for Go, Python, TypeScript, Vue.js (including Nuxt), Java, Rust, Kotlin, Swift, Ruby, and Elixir. All other files (config, docs, scripts) are also reviewed.

​
Models

Macroscope’s Auto-Tune system tests multiple model, prompt, and parameter combinations per language to find the best-performing config. An LLM-driven curator analyzes failures and proposes improvements each iteration. Learn more.