spencermarx/open-code-review

AI-powered multi-agent code review. Simulates a customizable team of Engineers performing code review with built-in discourse.

What it solves

Open Code Review (OCR) addresses the limitation of single-perspective AI code reviews. Instead of a single AI pass, it implements a multi-agent system where multiple independent reviewers—each with different personas and priorities—examine the code. This redundancy catches more issues, reduces false positives through a structured debate (discourse) between agents, and ensures code is evaluated against specific project requirements rather than just general best practices.

How it works

OCR uses an 8-phase workflow orchestrated by a "Tech Lead" agent:

  1. Context Discovery: Loads project standards from files like CLAUDE.md or .cursorrules.
  2. Change Analysis: Analyzes the git diff to understand the changes.
  3. Tech Lead Assessment: Summarizes risks and selects the appropriate reviewer team.
  4. Parallel Reviews: Multiple reviewer agents (using various LLMs) examine the code independently.
  5. Aggregation: Merges findings from the redundant reviewers.
  6. Discourse: Reviewers debate findings to validate concerns and challenge assumptions.
  7. Synthesis: Produces a prioritized, deduplicated final review.
  8. Presentation: Displays results in a dashboard or posts them to GitHub.

Who it’s for

Developers and teams who want to automate the first pass of code review to reduce the burden on human reviewers, specifically those using AI coding assistants like Claude Code, Cursor, or Windsurf.

Highlights

  • Multi-Agent Redundancy: Uses multiple independent reviewer instances to catch more bugs.
  • Customizable Personas: Includes 28 pre-built personas, including "Famous Engineers" (e.g., Martin Fowler, Kent Beck) and the ability to create custom or ephemeral reviewers.
  • Requirements-Aware: Can review code against specific spec files, proposals, or inline requirements.
  • Multi-Model Teams: Allows assigning different LLMs to different reviewers (e.g., a fast model for generalists, a deep model for specialists).
  • Code Review Maps: Generates structured navigation documents and Mermaid dependency graphs for large changesets (20+ files).
  • GitHub Integration: Can post reviews as a team synthesis or translate them into a natural "human voice" following Google's review guidelines.
  • Interactive Dashboard: A browser-based UI for triaging findings, managing teams, and tracking real-time progress.

Related

  • Project
    alibaba/open-code-review

    An AI-powered code review CLI tool that combines deterministic engineering with an agentic workflow to provide high-precision, token-efficient code analysis.

  • Project
    sunmh207/AI-Codereview-Gitlab

    An automated AI-driven code review tool for GitLab that uses LLMs to analyze code changes and post feedback directly to merge requests.

  • Dispatch
    Scaling AI Code Reviews: A Deep Dive into adamsreview for Claude Code

    An exploration of adamsreview, a multi-agent plugin for Claude Code designed to reduce false positives and catch deeper bugs through parallel validation and persistent state.

  • Project
    The-PR-Agent/pr-agent

    An open-source AI code review agent that automates PR descriptions, reviews, and improvements across multiple git platforms.

  • Project
    darrenhinde/OpenAgentsControl

    An AI coding framework that teaches agents your specific coding patterns to generate production-ready code that requires no refactoring, featuring human-guided approval gates and token-efficient context management.