vscode-mcp and mcp-server-vscode
These are competitors offering overlapping functionality—both provide MCP servers that expose VS Code's LSP diagnostics, code navigation, and editor capabilities to Claude/AI agents, with the primary difference being that tjx666's implementation emphasizes real-time LSP integration while malvex's emphasizes debugging tool access.
About vscode-mcp
tjx666/vscode-mcp
MCP server for Claude Code/VSCode/Cursor/Windsurf to use editor self functionality. ⚡ Get real-time LSP diagnostics, type information, and code navigation for AI coding agents without waiting for slow tsc/eslint checks.
Exposes eight MCP tools including symbol renaming with automatic import updates, comprehensive LSP information aggregation (definition, hover, signatures), and safe file operations—all backed by VSCode's language servers rather than subprocess execution. Operates as a monorepo with a VSCode extension bridge that maintains per-workspace socket connections to route MCP protocol requests into real-time LSP data, eliminating delays from external type-checking and linting commands. Supports Cursor, Claude Code, Windsurf, and Gemini CLI with granular tool filtering via environment variables or command-line arguments.
About mcp-server-vscode
malvex/mcp-server-vscode
Give Claude Code direct access to VS Code's code navigation, error checking, and debugging tools through the Model Context Protocol.
Exposes VS Code's language server capabilities (go-to-definition, find references, diagnostics, symbol search, call hierarchy) and debugger control (breakpoints, stepping, variable inspection, expression evaluation) as 25 MCP tools. Uses a three-tier architecture where a VS Code extension exposes an HTTP API on localhost:8991, a standalone MCP server bridges stdio-to-HTTP communication, and Claude Desktop connects via stdio transport—enabling seamless AI-assisted code navigation and interactive debugging workflows.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work