Vsc MCP

🚀 VSC-MCP Tool

This project offers tools to invoke Language Server Protocol (LSP) capabilities through the MCP (Model Context Protocol) tool. It enables AI agents to programmatically analyze and edit TypeScript/JavaScript code via the standardized MCP tool.

✨ Features

• Supports both single-file mode and project mode (detects the tsconfig.json file).
• Enables language-agnostic code operations via the LSP API.
• Fully compatible with the Model Context Protocol (MCP) SDK.

📦 Installation

Prerequisites

• Bun (for package management and building)
• Node.js (for running the server)

Setup

1. Clone the repository.
2. Install dependencies:

   bun install
   

🔒 Security Features

Path Restrictions

The VSC-MCP tool includes a security feature to prevent file operations from accessing unauthorized directories:

• ALLOWED_DIRECTORIES: An environment variable that specifies a comma-separated list of directories where file operations are allowed.

  Example:

  ALLOWED_DIRECTORIES=/home/user/projects,/tmp/workspace
  

• If this environment variable is not set, only operations in the current working directory are allowed by default.

• All file operations (read, write, edit) will check if the target path is within the allowed directories.

• Attempting to access files outside the allowed directories will result in an "Access Denied" error.

This feature helps prevent unauthorized access to sensitive files and directories on the system.

🔧 Technical Details

How It Works

• Start the TypeScript language server in standard input/output mode.
• Send LSP's initialize and didOpen notifications.
• Use workspace/symbol and textDocument/documentSymbol requests to find symbols.
• When symbols cannot be found, fall back to textDocument/implementation for additional symbol discovery.
• Apply edit operations directly to files using the applyWorkspaceEdit tool.
• Return the operation results to the caller.

💻 Usage Examples

Available Tools

The following tools are currently available:

• editSymbol: Edit symbols (functions, classes, methods, etc.) in a file by name and type via LSP.

  • Supported symbol types include: functions, methods, classes, interfaces, variables, constants, properties, fields.
  • Use LSP's workspace/symbol and textDocument/documentSymbol requests to find symbols.
  • Return structured information such as file path and location.

• readSymbol: Read symbols (functions, classes, methods, etc.) in a file by name and type via LSP.

  • Supported symbol types include: functions, methods, classes, interfaces, variables, constants, properties, fields.
  • Use LSP's workspace/symbol and textDocument/documentSymbol requests to find symbols.
  • Return structured information such as file path and location.

• findReferences: Find all references to symbols (functions, classes, etc.) in a file via LSP.

  • Supported symbol types include: functions, methods, classes, interfaces, variables, constants, properties, fields.
  • Use LSP's textDocument/references request to find all symbol references.
  • Return a list of references, including file path and location.
  • Assist in code analysis, refactoring, and understanding code usage.

• getDefinition: Get the definition of symbols (functions, classes, etc.) via LSP.

  • Supported symbol types include: functions, methods, classes, interfaces, variables, constants, properties, fields.
  • Use LSP's textDocument/definition request to get symbol definitions.
  • Return structured information such as file path and location.

• getDocumentation: Get the documentation comments of symbols (functions, classes, etc.) via LSP.

  • Supported symbol types include: functions, methods, classes, interfaces, variables, constants, properties, fields.
  • Use LSP's textDocument/hover and textDocument/documentSymbol requests to get documentation information.
  • Return structured information such as documentation content and location.

• getCodeIssues: Check code issues via LSP.

  • Use LSP's textDocument/formatting and textDocument/typeDefinition requests to check code format and type definitions.
  • Return structured information such as errors, warnings, and locations.