MCP Chunk Editor

What is MCP Chunk Editor?

MCP Chunk Editor is a Model Context Protocol server that enables LLMs to edit text files using semantically meaningful chunks (functions, classes, methods) rather than line numbers. It uses Universal CTags to parse files into chunks and provides immediate application of edits with preview. It is designed for AI-assisted code editing workflows.

How to use MCP Chunk Editor?

Install via pip install mcp-chunk-editor or run directly with uvx mcp-chunk-editor. Configure it in Claude.app by adding a JSON entry under mcpServers using the uvx, docker, or python command. Optionally set a custom --max-chunk-size argument (default 100 lines) to control chunk sizes.

Key features of MCP Chunk Editor

• Semantic chunking using Universal CTags
• Immediate change application with preview generation
• Simple undo for the most recent change per file
• Optimized for LLM token usage with incremental updates

Use cases of MCP Chunk Editor

• An LLM reads a file’s function-level chunks, then replaces a specific method’s content
• An LLM creates a new file with create_file and later deletes it with delete_file
• An LLM edits a codebase chunk by chunk, undoing a change if the preview shows an error

FAQ from MCP Chunk Editor

What Python version is required?

Python 3.10 or 3.11 is required; Python 3.12+ is not supported yet due to python-ctags3 compatibility.

Does MCP Chunk Editor support undo for multiple changes?

It supports undoing the most recent change per file, with one level of undo.

What external dependency does it need?

Universal CTags must be installed (e.g., apt-get install universal-ctags on Debian/Ubuntu).

Can I adjust the maximum chunk size?

Yes, pass the --max-chunk-size argument to set a custom maximum chunk size (default is 100 lines).