mcp-performance

A Gemini CLI extension and MCP server to analyze web page performance

Gemini CLI Extension

This project is intended to be installed as a Gemini CLI extension. This will include both the MCP server and commands for consistent AI responses. To install it, use the following command:

gemini extensions install https://github.com/mcclatchy/mcp-performance

After installing, you will have access to a performance MCP server as well as commands that we have created to assist you. Commands begin with the / character, such as /performance-analysis, and will be visible in the client. These commands will automatically pull performance data using the MCP server. If you would like to pull performance data for a different conversation, you can initiate the MCP tool directly in a prompt. This direct invocation pattern works consistently but is not strictly necessary when using the Gemini CLI commands:

`performance-entries` --url="https://www.coralspringsflnews.com" --mark="flex first rendered"

MCP Server

This project can also be run as a standalone MCP server. This is useful for developers who want to use this tool with other MCP clients, such as mcp-inspector, or other LLM clients like Claude. To run the server for these clients, use the following command:

npx github:mcclatchy/mcp-performance

For local development and testing with the Gemini CLI, please refer to the "Local Development" section and use npm start.

Note: you do not need to use the npx command if using Gemini CLI.

Local Development

For developers working on the mcp-performance project, follow these steps:

Installation

This project requires Node.js and npm. To install the development dependencies, run the following command:

npm install

Running the MCP Server Locally

To run the MCP server for local development, use the following command:

npm start

This will start the MCP server and make it available for testing with the Gemini CLI or other MCP clients.

Using Performance Marks

The performance-entries tool can wait for a specific performance.mark to be set on the page before collecting performance metrics. This is particularly useful for measuring custom events or ensuring that a specific part of the application has loaded or executed.

To use this feature, provide the mark parameter with the name of the performance mark you wish to wait for. For example, if your webpage sets a mark like performance.mark('myCustomEvent'), you can instruct the tool to wait for it.

Example (when used with Gemini CLI):

gemini performance-entries --url "https://example.com" --mark "myCustomEvent"