Diffchunk

Provides navigation and targeted analysis of large diffs to help LLMs pinpoint relevant changes.

other

8

GitHub Stars

other

Language

5 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "peteretelej-diffchunk": {
      "command": "uvx",
      "args": [
        "--from",
        "diffchunk",
        "diffchunk-mcp"
      ]
    }
  }
}

You can use the DiffChunk MCP Server to let large diff files be navigated efficiently by guiding your AI assistant to relevant changes. It reduces token waste by jumping directly to meaningful chunks instead of processing entire diffs, making large-scale reviews practical.

How to use

Configure your MCP client to connect to the DiffChunk MCP Server, then use its navigation tools to explore large diffs. The server exposes four core tools that you can invoke during analysis: load_diff to parse a diff with optional settings, list_chunks to get an overview of the diff structure, get_chunk to retrieve a specific chunk, and find_chunks_for_files to locate chunks by file patterns. Typical workflows include obtaining a high-level view of the diff, selecting relevant Python or JavaScript chunks, and inspecting specific sections without loading the entire file into context.

Usage scenarios you can perform with your AI assistant include branch comparisons, code reviews, and change analysis. For example, you can ask the assistant to summarize all changes in a branch, review a feature branch for security considerations, or identify API changes affecting a mobile app. You can also apply these tools to analyze changes at specific paths or patterns, such as focusing on Python files or database migrations.

How to install

Prerequisites include having uv, a fast Python package manager, installed so you can run the MCP server. Ensure you have Python available on your system before proceeding.

Step 1: Install uv to obtain the uvx command.

Step 2: Verify uvx is available.

Step 3: Run the DiffChunk MCP Server using the provided MCP entry configuration.

Configuration and usage notes

The MCP server auto-loads diffs with sensible defaults to help you start quickly. You can customize behavior by loading a diff with specific options such as maximum chunk size or include/exclude patterns. The recommended sequence for analyzing large changesets is: create the diff, list chunks for an overview, locate relevant chunks by files, fetch individual chunks as needed, then clean up temporary files.

Example workflow patterns you can run against a diff file include loading a large diff, listing chunks to understand the scope, filtering to Python files, and retrieving a specific chunk to review in detail.

Notes on formats and performance

The server supports common diff formats such as Git diff output and unified diffs, even when multiple files are included in a single diff. It is designed to handle very large diffs efficiently by streaming data and loading only the needed chunks into context.

Performance is optimized for very large changesets, with memory-efficient streaming and automatic reloading on file changes to keep analyses up to date.

Available tools

load_diff

Parse a diff file with optional settings to customize how chunks are produced.

list_chunks

Show a high-level overview of the diff, including how many chunks and which files are involved.

get_chunk

Retrieve content for a specific chunk to analyze a section without loading the entire diff.

find_chunks_for_files

Locate chunk indices that correspond to files matching given patterns to focus the review.