mcp-server-chatgpt-mac-app
# mcp-server-chatgpt-mac-app
This is an MCP server that allows you to use the ChatGPT macOS app via MCP.
The original version was written in Python as [mcp-server-chatgpt-app](https://github.com/cdpath/mcp-server-chatgpt-app),
and this project is a TypeScript implementation of it.
## Prerequisite
- [ChatGPT macOS app](https://openai.com/chatgpt/download/)
- [Install the "Ask ChatGPT on Mac" shortcuts](https://www.icloud.com/shortcuts/f8f249e8fce7450291f385874c64c6eb)
- Node.js
## Using npx
The MCP server can be used via npx,
and you can choose to use it either globally on your macOS system or within any local project.
### Global Usage
If you want to use it globally, run the following command in your CLI.
```bash
claude mcp add chatGPTOnMac -s user \
-- npx mcp-server-chatgpt-mac-app
```
### Local Project Usage
If you want to use it in any local directory,
add the following to your `.mcp.json`.
```json
{
"mcpServers": {
"chatGPTOnMac": {
"command": "npx",
"args": ["mcp-server-chatgpt-mac-app"],
"cwd": ".",
"env": {}
}
}
}
```
### How to check if the MCP server has been added
To verify that the MCP server has been added, run the following command:
```bash
claude mcp list
```
If the output looks like the following, everything is set up correctly.
```bash
$ claude mcp list
chatGPTOnMac: npx mcp-server-chatgpt-mac-app
```
## Local Development
If you want to improve this project locally.
### build
```bash
git clone https://github.com/yimajo/mcp-server-chatgpt-mac-app.git
cd mcp-server-chatgpt-mac-app
pnpm install
pnpm build
```
TDQS
Scored across 1 tool
With only one tool, there is no risk of confusion. The tool's purpose is clearly defined and distinct from any other possible tools.
Since there is only one tool, naming consistency is not applicable; a single tool cannot have inconsistencies. The name 'ask_via_shortcuts_cli' is descriptive and follows a reasonable pattern.
A single tool is at the lower end of what is reasonable. For a server targeting ChatGPT Mac app interaction, one tool feels thin but may be justified if the intended use case is very narrow.
The tool set lacks many expected operations for interacting with an AI assistant, such as handling conversations, retrieving responses in different formats, or managing configurations. The surface is significantly incomplete for its domain.