MCP-MultilspyLSP

An MCP server that exposes Language Server Protocol (LSP) capabilities through multilspy.

What is this?

This project provides an MCP (Model Context Protocol) server that gives language models access to language intelligence features from the Language Server Protocol (LSP). It uses multilspy, a Python library that simplifies interfacing with various language servers.

Features

Language Intelligence: Get code completions, find definitions, references, and more
Multi-language Support: Works with multiple programming languages including Python, Java, TypeScript, and more
Easy Integration: Works with any MCP-compatible client like Claude Desktop
Static Analysis: Leverage the power of language servers for code understanding

Installation

pip install mcp-multilspy

Usage

Start the server directly:

mcp-multilspy

Or install it in Claude Desktop:

mcp install mcp-multilspy

Supported Languages

Java (Eclipse JDTLS)
Python (jedi-language-server)
Rust (Rust Analyzer)
C# (OmniSharp/RazorSharp)
TypeScript (TypeScriptLanguageServer)
JavaScript (TypeScriptLanguageServer)
Go (gopls)
Dart (Dart Language Server)
Ruby (Solargraph)

Example

Here's how to use this server with an MCP client:

# Initialize a TypeScript language server session
session = await initialize_language_server(
    session_id="ts-session", 
    project_root="/path/to/project",
    language="typescript"
)

# Find where a symbol is defined
definitions = await request_definition(
    session_id="ts-session",
    file_path="src/index.ts",
    line=10,  # 0-indexed
    column=15  # 0-indexed
)

# Find all references to a symbol
references = await request_references(
    session_id="ts-session",
    file_path="src/index.ts",
    line=10,
    column=15
)

# Get code completion suggestions
completions = await request_completions(
    session_id="ts-session",
    file_path="src/index.ts", 
    line=10,
    column=15
)

# Get hover information
hover_info = await request_hover(
    session_id="ts-session",
    file_path="src/index.ts",
    line=10,
    column=15
)

# Get document symbols
symbols = await request_document_symbols(
    session_id="ts-session",
    file_path="src/index.ts"
)

# Clean up when done
await shutdown_language_server(session_id="ts-session")

Requirements

Python 3.12 or higher
Language-specific requirements:
- For Java: JDK 17.0.6 or higher
- For JavaScript/TypeScript: Node.js v18.16.0 or higher

License

MIT

Integrations