How to Set Up the code-review-graph Repository Locally: Complete Installation Guide

You can set up code-review-graph by cloning the repository, installing the CLI via pip or uv, running code-review-graph install to auto-configure MCP integrations, and executing code-review-graph build to generate the initial SQLite knowledge graph.

The code-review-graph tool creates a persistent, incrementally-updated knowledge graph of your codebase using Tree-sitter parsing and SQLite storage. Setting up the repository locally gives you a local-first AI code review system with no telemetry, exposing functionality through both a command-line interface and an MCP server for integration with AI coding assistants.

Prerequisites

Before beginning the local setup, ensure you have Python 3.x installed on your system. The tool supports installation via pip, pipx, or uv, though a virtual environment is strongly recommended to isolate dependencies.

Step-by-Step Local Setup

Clone the Repository

First, download the source code from GitHub and navigate into the project directory:

git clone https://github.com/tirth8205/code-review-graph.git
cd code-review-graph

This provides access to the core parser logic in code_review_graph/parser.py and the documentation in docs/architecture.md.

Create a Python Virtual Environment

Isolate your dependencies to avoid conflicts with system packages:

python3 -m venv .venv
source .venv/bin/activate

On Windows, use .venv\Scripts\activate instead.

Install the CLI Tool

Choose one of three installation methods depending on your workflow:

• Pip (standard): pip install code-review-graph
• Pipx (global CLI isolation): pipx install code-review-graph
• uv (fast Rust-based installer): uv tool install . --force

All methods install the code-review-graph command-line interface. According to the source, the install sub-command will later handle platform-specific hook configurations.

Configure Platform Integrations

Auto-detect your AI coding tools and write the correct MCP configuration files:

code-review-graph install

This command inspects your environment for supported editors and AI assistants, then installs the appropriate hooks and MCP server configurations needed for integration.

Build the Initial Knowledge Graph

Parse your entire repository and create the SQLite graph database:

code-review-graph build

This first build parses all tracked files using the Tree-sitter grammar engine implemented in code_review_graph/parser.py, extracting nodes (files, classes, functions) and edges (calls, imports, inheritance). For a typical 500-file project, expect approximately 10 seconds for completion. The graph persists to .code-review-graph/graph.db.

Enable Live Updates (Optional)

Start a file watcher to keep the graph synchronized automatically:

code-review-graph watch

The incremental engine detects changed files via Git or SVN, re-parsing only modified files plus their dependents rather than rebuilding the entire graph.

Verify the Installation

Confirm the graph is operational and view statistics:

code-review-graph status
code-review-graph detect-changes --brief

These commands display graph statistics and a token-savings panel, confirming that the SQLite store at .code-review-graph/graph.db is properly initialized and accessible.

Understanding the Core Architecture

The local setup creates three primary components on your machine:

• Parser – Uses Tree-sitter grammars in code_review_graph/parser.py to extract AST-based structural nodes and edges from every tracked file.
• Graph Store – A SQLite database located at .code-review-graph/graph.db that persists nodes, edges, metadata, and optional embeddings.
• Incremental Engine – Detects file changes and updates only the affected portions of the graph, making subsequent builds nearly instantaneous.

Key Configuration Files

After setup, several files govern the tool's behavior:

File	Purpose
.code-review-graph/graph.db	The SQLite database containing the persisted knowledge graph (auto-generated after first build).
.code-review-graphignore	Custom ignore patterns for indexing; uses the same syntax as .gitignore. Create this in your repository root as needed.
docs/USAGE.md	Detailed documentation for CLI commands and platform-specific options.
docs/architecture.md	System overview and data-flow diagrams describing how the parser and incremental engine interact.

Summary

• Clone the repository from https://github.com/tirth8205/code-review-graph.git to access the source.
• Install using pip, pipx, or uv to obtain the code-review-graph CLI.
• Configure MCP integrations automatically with code-review-graph install.
• Build the initial SQLite graph using code-review-graph build, which leverages code_review_graph/parser.py for AST extraction.
• Maintain freshness via code-review-graph watch for incremental updates or manual rebuilds.
• Customize indexing behavior using .code-review-graphignore files in your project root.

Frequently Asked Questions

What Python version is required to run code-review-graph locally?

The repository requires Python 3.x or higher. While specific minimum versions depend on dependencies like Tree-sitter bindings, any modern Python 3 installation should suffice. Using a virtual environment prevents conflicts with system packages.

Where does code-review-graph store the knowledge graph data?

All data persists locally in a SQLite database at .code-review-graph/graph.db. This local-first approach ensures no telemetry is transmitted unless you explicitly opt-in to cloud embeddings. The database stores nodes (files, classes, functions), edges (calls, imports), and metadata.

How do I integrate code-review-graph with my AI coding assistant?

Run code-review-graph install after installation. This command auto-detects supported AI tools and editors on your system, then writes the necessary MCP (Model Context Protocol) configuration files. Once configured, you can ask your MCP client: "Review my recent changes with risk scoring" to leverage the graph for contextual code review.

Can I exclude specific files from the knowledge graph indexing?

Yes. Create a .code-review-graphignore file in your repository root using standard .gitignore syntax. The parser respects these patterns during both full builds and incremental updates, allowing you to exclude generated files, dependencies, or test data from the graph.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}