AI

In the post on Wolfram Cloud MCP, we are using the following prompt to create and display the plot in Ghostty. Use the Wolfram to generate a plot of Plot[Sin[x], {x, 0, 10}]. Take the resulting output URL or image payload and display it inline using your image-view tool. Clearly, it is too tedious to write a prompt this long every time, and we can create a skill to simplify it. A skill is essentially a series instruction for AI agents to perform multiple-step workflow. ...

I have tested the local MCP server on Mac, and today I am testing Wolfram Cloud MCP server in ChromeOS Flex, currently it free for the personal use. To use MCP in pi.dev, install npm:pi-mcp-extension first. Here is the minimum configuration for mcp.json. 1{ 2 "settings": { 3 "toolPrefix": "mcp", 4 "requestTimeoutMs": 30000, 5 "maxRetries": 5 6 }, 7 "mcpServers": { 8 "wolfram": { 9 "transport": "streamable-http", 10 "url": "https://agenttools.wolfram.com/mcp", 11 "lifecycle": "eager" 12 } 13 } 14} Ghostty terminal supports the Kitty graphics protocol to display inline images directly within the terminal, it will be nice if we can display the plots generated from the Wolfram Cloud. ...

Note: this is simply a test of the Wolfram MCP integration with AI — nothing more. With the Wolfram MCP server, we can attempt something more complex than plain coding. ...

In the past, I occasionally tried to get AI to test novel visualization ideas in Wolfram. No matter which AI I used — Claude, ChatGPT, or Gemini — the generated code was barely usable at best. Now, with Wolfram MCP, the AI model should be able to produce much better code. I saw 507 Mechanical Movements on Hacker News. Let’s have some fun with it. ...

Wolfram Local MCP Since Wolfram (or Mathematica) 15.0, Wolfram Local MCP works out of the box. Wolfram had already configured the MCP server for my Claude Desktop, so I reused that configuration file for pi.dev. ...