mcp-code-execution

What problem does it solve?

This Skill dramatically reduces the token cost of executing Serena MCP tools by running them as TypeScript code. It ensures you only load the necessary tools into memory, process data locally within your script, and report only the final results to Claude, saving time and avoiding context bloat.

Core Features & Use Cases

• On-demand tool loading: Import only the Serena MCP tools you actually need, when you need them.
• Token-efficient workflows: Process large datasets entirely in-code and return a concise result, minimizing LLM context.
• Modular, multi-tool workflows: Seamlessly compose tasks like listing files, inspecting symbols, and refactoring, while keeping intermediate data private to your script.
• Use Case: You have 1000+ source files. Use this Skill to analyze and transform them with Serena MCP, while the final summary is sent to the LLM for decision-making.

Quick Start

From your project, place the MCP code execution script in the Skill’s context and run it to perform deterministic steps. Then call the script via bun run, for example: bun run src/your-script.ts