Is there a way to test the changes before applying them?

Yes, the skill defaults to a dry-run mode that shows the optimization plan and estimated savings before you confirm the file write.

Why does it suggest a restart after optimization?

Claude Code loads MCP configurations when the session starts, so a restart is required for the new local settings to take effect.

What is the difference between optimization and conversion?

Optimization keeps your MCP servers but limits when they load. Conversion (mcp-to-skills) replaces the server entirely with a portable skill.

How do I undo the optimization?

The process is fully reversible; simply delete the .mcp.json file in your project directory to return to your global configuration.

Does this skill modify my global Claude settings?

No, it specifically creates a project-local .mcp.json file. It never modifies your global ~/.claude.json configuration.

MCP Project Optimizer

Name: MCP Project Optimizer
Author: choam2426

bychoam2426

•

Productivity & Workflow

Reduces Claude Code token waste by generating project-specific MCP configurations that scope tools to the current workspace.

The MCP Project Optimizer is a high-efficiency utility designed to minimize idle token costs and streamline the Claude Code environment. By analyzing session history, the skill identifies which MCP servers are relevant to your current project and generates a minimal, local .mcp.json configuration. This approach allows developers to maintain their MCP ecosystem while preventing unnecessary servers from consuming context window space. It provides a low-risk, fully reversible optimization path that includes detailed token savings reports and a dry-run mode for safe analysis before any changes are applied.

Key Features

01Generates project-local .mcp.json overrides to limit tool scope

02Dry-run capability to preview optimization plans without writing files

03Automated session analysis to identify high-waste MCP servers

04Non-destructive workflow that preserves global configuration files

058 GitHub stars

06Detailed reporting with estimated token savings and server status

Use Cases

01Reducing recurring token costs in Claude Code by offloading unused global servers

02Auditing MCP server performance to identify the best candidates for optimization

03Scoping specialized development tools to specific repositories for cleaner context

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add choam2426/mcp-optimizer mcp-optimize

For use in Claude.ai and ChatGPT

Key Features

01Generates project-local .mcp.json overrides to limit tool scope

02Dry-run capability to preview optimization plans without writing files

03Automated session analysis to identify high-waste MCP servers

04Non-destructive workflow that preserves global configuration files

058 GitHub stars

06Detailed reporting with estimated token savings and server status

Use Cases

01Reducing recurring token costs in Claude Code by offloading unused global servers

02Auditing MCP server performance to identify the best candidates for optimization

03Scoping specialized development tools to specific repositories for cleaner context

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add choam2426/mcp-optimizer mcp-optimize

For use in Claude.ai and ChatGPT