@first-to-fly/code-cleanup
A Bun TypeScript project that provides a Model Context Protocol (MCP) server for cleaning up code files using Google's Generative AI. It backs up original files in a .stash directory and allows users to optionally define custom system prompts for the cleanup process.
Code Cleanup MCP Server
A Bun TypeScript project that provides a Model Context Protocol (MCP) server for cleaning up code files using Google's Generative AI. It backs up original files in a .stash directory and allows users to optionally define custom system prompts for the cleanup process.
Features
- Cleans up code files with a professional software engineering focus (removes unused imports, ensures consistent formatting, etc.).
- Supports custom system prompts for flexible cleanup instructions.
- Backs up original files in a
.stashdirectory before modification. - Provides tools to manage the stash directory.
- Built with strong TypeScript typing for reliability.
Installation
For Roo Code
{
"mcpServers": {
"code-cleanup": {
"type": "stdio",
"timeout": 300,
"command": "npx",
"args": ["--yes", "@first-to-fly/code-cleanup"],
"env": {
"CODEBASE_PATH": "__PWD__",
"GOOGLE_API_KEY": "__CODEBASE_GOOGLE_API_KEY__",
<!-- "SYSTEM_INSTRUCTION": "This is the instruction to clean up the code.", -->
<!-- "MODEL": "gemini-1.5-pro by default. Feel free to use different one." -->
},
"alwaysAllow": ["cleanup_code_files"],
"disabled": false
}
}
}
Default SYSTEM_INSTRUCTION
Clean up the provided code like a professional software engineer, focusing on:
* Removing unused imports, variables, and redundant one line comments (retain only meaningful comment and documentation).
* Ensuring consistent naming and formatting according to language best practices.
* Simplifying minor inefficiencies (e.g., redundant calculations) *without* altering the core logic.
* Removing unnecessary whitespace while preserving single-line breaks between logical blocks of code.
Crucially, *do not* change the code's original logic, variable names (unless obviously incorrect style), or overall functionality. Do not add any new comments except to clarify existing deprecated code. Do not rewrite or restructure major sections of code.
Output *only* the cleaned, raw code, with proper indentation and formatting. Do not include any introductory phrases, explanations, annotations, or markdown formatting (backticks or otherwise). The output should be the code itself, ready to be copied and pasted.
Example:
**Not like this:**
\\`\\`\\`javascript
console.log("hello");
\\`\\`\\`
**Like this:**
console.log("hello")
---
Provide back raw code.
Recommend MCP Servers 💡
ayushps1/remote-mcp-server
A remote Model Context Protocol (MCP) server designed to run on Cloudflare Workers, providing tools (like a math tool) accessible via SSE with OAuth login.
@charlesmuchene/pref-editor-mcp-server
Edit Android app user preferences in real-time using natural language commands via MCP protocol
Office-PowerPoint-MCP-Server
A MCP server for PowerPoint manipulation using python-pptx, providing tools for creating, editing, and manipulating PowerPoint presentations through the MCP protocol.
awslabs.nova-canvas-mcp-server
An MCP server that enables text-based and color-guided image generation using Amazon Nova Canvas, supporting customizable dimensions, quality, and multiple image outputs, with secure AWS authentication and workspace integration.
gibwork-mcp
An MCP server for managing tasks (create, retrieve, search) on the GibWork platform via the Model Context Protocol
lean-lsp-mcp
MCP server that allows agentic interaction with the Lean theorem prover via the Language Server Protocol using leanclient