Connect KevsRobots to your AI assistant

Ask an AI assistant about a robot build and it will often half-remember something, or make it up entirely. This lets it search this site properly instead - so its answers come from the actual tutorials, with links you can click and check.

It uses MCP (Model Context Protocol), an open standard for connecting AI apps to outside tools. Setup is one line, it is free, and it is read-only - there is no account and nothing to sign up for.

The endpoint

https://search.kevsrobots.com/mcp

That is all most apps need.

Claude Code

In your terminal:

claude mcp add --transport http kevsrobots https://search.kevsrobots.com/mcp

Then ask it something like “search kevsrobots for how to wire a DRV8833”.

Claude Desktop

1. Open Settings and go to Connectors
2. Choose Add custom connector
3. Name it KevsRobots and paste in https://search.kevsrobots.com/mcp
4. Save, then start a new chat

Any other MCP client

Most other apps take a JSON config. The endpoint uses the streamable HTTP transport:

{
  "mcpServers": {
    "kevsrobots": {
      "type": "http",
      "url": "https://search.kevsrobots.com/mcp"
    }
  }
}

Some clients still expect the older command/args shape for local servers. If yours has no HTTP option, it will need a bridge such as mcp-remote - check your app’s own MCP documentation.

What your assistant can do with it

Two tools become available:

The search covers everything on the site: courses and lessons, blog posts, project write-ups, robot builds, hardware reviews and how-it-works explainers.

Things worth trying once it is connected:

• “Search kevsrobots for a beginner MicroPython course and summarise what it covers”
• “What does kevsrobots say about wiring a PCA9685 to a Pico?”
• “Find the parts list for the SMARS robot”

Good to know

It is read-only. The tools search and read public pages. Nothing writes anything, and no account or key is involved.

Answers should carry links. The tools return a URL with every result, and assistants are asked to cite them. If you get an answer with no link, be sceptical - it may not have used the search at all.

It searches, it does not think. The assistant does the reasoning; this just gives it the right pages to reason from. If the site does not cover something, the honest answer is that there is nothing here about it.

It cannot tell you a page is wrong. If a tutorial has a mistake in it, the assistant will faithfully repeat that mistake with a citation. The link is there so you can check - use it.

Trouble

• Assistant does not use the tools - ask explicitly: “search kevsrobots for…“. Some apps only reach for a tool when the request clearly calls for it.
• Connection fails - check the URL has no trailing slash and starts with https.
• No results - try fewer words, or drop any filters. A full question generally works better than one keyword.

If something is broken, open an issue and let me know.