MergeGuard

Cross-PR intelligence for the agentic coding era.

MergeGuard detects conflicts between open pull requests before they become merge headaches. It analyzes overlapping code changes across PRs using AST-level understanding, computes risk scores, and integrates seamlessly as a GitHub Action or CLI tool.

Why MergeGuard?

As AI coding agents (Cursor, Copilot, Devin, Claude Code) generate more PRs in parallel, the likelihood of cross-PR conflicts increases dramatically. Traditional CI only checks a single PR against the base branch — it can't see that two PRs are about to break each other.

MergeGuard fills this gap by:

• Detecting hard conflicts — two PRs modify the same function body
• Catching interface conflicts — one PR changes a function signature while another PR calls it
• Identifying behavioral conflicts — incompatible logic changes in the same module
• Flagging duplications — two PRs implementing the same feature independently
• Detecting regressions — a PR re-introduces something recently removed
• Computing risk scores — composite scoring based on conflict severity, blast radius, code churn, and AI attribution

Quick Start

As a GitHub Action

# .github/workflows/mergeguard.yml
name: MergeGuard
on:
  pull_request:
    types: [opened, synchronize, reopened]

jobs:
  analyze:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: Vansh2795/mergeguard@v0.1
        with:
          github-token: ${{ secrets.GITHUB_TOKEN }}

As a CLI

# Install from PyPI
pip install py-mergeguard

# Analyze a specific PR
mergeguard analyze --repo owner/repo --pr 42 --token $GITHUB_TOKEN

# Show collision map of all open PRs
mergeguard map --repo owner/repo --token $GITHUB_TOKEN

# Risk dashboard for all open PRs
mergeguard dashboard --repo owner/repo --token $GITHUB_TOKEN

How It Works

1. Fetch — Retrieves all open PRs and their diffs from GitHub
2. Parse — Uses Tree-sitter to build AST-level understanding of changed code
3. Detect — Compares every pair of open PRs for overlapping changes
4. Classify — Categorizes conflicts (hard, interface, behavioral, duplication, regression)
5. Score — Computes a composite risk score (0-100) for each PR
6. Report — Posts actionable comments on PRs or displays results in the terminal

Configuration

Create a .mergeguard.yml in your repository root:

risk_threshold: 50        # Only comment if risk score > 50
check_regressions: true   # Detect regressions against recent merges
max_open_prs: 30          # Performance limit
llm_enabled: false        # Optional AI-powered semantic analysis
ignored_paths:
  - "*.lock"
  - "package-lock.json"

See Configuration Guide for all options.

Documentation

• Getting Started
• How It Works
• Configuration
• Architecture
• Contributing
• FAQ

Development

# Clone and setup
git clone https://github.com/Vansh2795/mergeguard.git
cd mergeguard
uv sync --dev

# Run tests
uv run pytest

# Lint
uv run ruff check src/ tests/
uv run mypy src/

License

MIT — see LICENSE.