Purpose
MCP AI Server is an MCP server extension for Visual Studio, designed to give AI coding assistants the same code understanding capabilities as the IDE. Traditional AI coding tools (such as Claude Code, Codex CLI, Gemini CLI, Cursor, Copilot, etc.) work at the file system level and can only read text, run grep, and execute builds; they cannot understand code structure the way Visual Studio does. This extension exposes IntelliSense-level intelligence and the Visual Studio debugger as MCP tools through the Model Context Protocol, enabling AI assistants to perform semantic understanding rather than text matching.
Main Capabilities
The extension provides a total of 41 tools, divided into two categories: stable tools and preview debugging tools.
Stable tools (22):
- Semantic navigation (powered by Roslyn): supports semantic-level navigation similar to F12 (Go to Definition) and Shift+F12 (Find All References).
- Code understanding (powered by Roslyn): helps AI understand code structure and meaning.
- Code analysis (powered by Roslyn): performs semantic analysis on code.
- Refactoring (powered by Roslyn): supports refactoring operations such as safe rename.
- Projects and build: involves project management and build execution.
- Editor integration: interacts with the Visual Studio editor.
Debugging tools (19, preview):
- Debug control (10 tools)
- Debug inspection (5 tools)
- Breakpoint management (4 tools) AI assistants can debug .NET code at runtime through the Visual Studio debugger, including setting breakpoints, stepping, inspecting variables, and attaching to Docker containers and WSL processes.
How to Use
- Open a .NET solution in Visual Studio.
- Configure the port in the MCP Server settings (default 3010); the path is Tools -> MCP Server -> Global Settings.
- Start the MCP server in Visual Studio: Tools -> MCP Server -> Start Server.
- Add the serverto the MCP client.
- Ask the AI semantic questions about the code.
MCP client configuration supports multiple approaches:
- Clients with native HTTP support (such as Claude Desktop and Claude Code) can be configured directly.
- Clients that do not support HTTP can connect through the mcp-remote proxy.
- Codex CLI is configured using the TOML format.
It is recommended to add instructions to the project's AI configuration file to ensure that the AI automatically prioritizes the use of MCP tools.
Multi-solution scenarios: If the AI needs to understand dependent libraries, you can clone the source code from GitHub and open it in a second Visual Studio instance. Each instance runs its own MCP server on a configurable port, and the AI can connect to all instances simultaneously to trace calls across code and library code, find usage patterns, and understand inheritance relationships.
Applicable Scenarios
- .NET development that requires an AI assistant to perform semantic-level code navigation, symbol search, inheritance tree, and call graph analysis.
- Automated workflows that require AI to perform safe rename, build, and test.
- Scenarios that require AI to debug .NET code at runtime, including Docker and WSL environments.
- Developers using any MCP-compatible AI tool (Claude Code, Codex CLI, Gemini CLI, OpenCode, Goose, Aider, Claude Desktop, Cursor, Windsurf, VS Code + Copilot, Cline, Continue, etc.).
Limitations
- System requirements: Visual Studio 2022 (17.13+) or Visual Studio 2026, Windows (amd64 or arm64).
- The debugging tools are in preview.
- An MCP-compatible AI tool is required.
- Port configuration changes take effect the next time the server starts.
- Information such as fees and pricing is not currently provided in the materials.
