Assistant MCP Server

What is Assistant MCP Server?

An MCP server that provides tools for retrieving project architecture information, searching tasks, and generating optimized prompts for AI models. It is designed for developers working on frontend applications who need to integrate structured project data into AI workflows.

How to use Assistant MCP Server?

Install dependencies with yarn install, then build with yarn build. Add a tools.json file defining the available tools. To run locally, configure an MCP client with the command npx tsx /path/to/src/index.ts and set the TOOLS_PATH environment variable to the path of tools.json.

Key features of Assistant MCP Server

• Provides three pre‑defined tools: architecture_info, search_tasks, and optimize_prompt.
• Uses a tools.json configuration file for tool definitions.
• Supports file‑based plugins (e.g., reading architecture info and tasks from local files).
• Includes a promptOptimizer plugin for generating final structured AI prompts.
• Runs locally via npx tsx for easy development and testing.

Use cases of Assistant MCP Server

• Automatically retrieve frontend project architecture before implementing new tasks.
• Collect project tasks from a tasks.txt file and guide AI to execute them according to architecture.
• Generate a final, structured prompt by combining multiple context sections and instructions for an AI model.

FAQ from Assistant MCP Server

What dependencies are required?

The server requires Node.js and TypeScript (via tsx). Dependencies are installed with yarn install.

How do I configure the tools?

Create a tools.json file at the project root with an array of tool objects. Each tool specifies its name, description, input schema, and plugin configuration (e.g., file or promptOptimizer).

Where is the data stored?

Tool data resides in local files as specified in the plugin arguments (e.g., architecture.md and tasks.txt). The server reads these files at runtime.

What transport does Assistant MCP Server use?

The server uses standard MCP stdio transport – it is invoked as a command‑line process via npx tsx.

Is there any authentication or security mechanism?

The README does not mention any authentication. The server runs locally and relies on file‑system access to the configured paths.