mcp

Give your AI agent live access to the Brick Discovery catalog — no scraping, no stale answers. Connect once, then just ask.

What you get

A connected agent can search and browse the full Brick Discovery catalog — tools, apps, communities, categories, and full tool details — and ask what changed this week. Results are served from the same weekly-refreshed snapshot that powers the site, so answers reflect what's actually published, not a stale scrape.

Any product link an agent hands back comes through a /go/<slug> redirect, the same attribution links used across the site — not raw URLs.

Try asking

Once connected, your agent can answer things like:

"Find me a free web tool for planning a LEGO MOC's parts list"
"What's new in the AFOL tool scene this week?"
"Which LEGO YouTube channels cover Technic?"

The endpoint

https://www.brickdiscovery.com/mcp — free, read-only, no account, no API key. It's a stateless streamable-HTTP MCP server exposing search and lookups over the published tool catalog. Details on the protocol and each tool's request/response shape are in docs/mcp.md.

Please use it fairly — reasonable request volumes for personal or assistant use. See our terms & conditions for the full fair-use policy.

Set it up

Claude Desktop

Settings → Connectors → Add custom connector, then use this URL:

Claude Desktop connector URL

https://www.brickdiscovery.com/mcp

Claude Code

claude mcp add --transport http brickdiscovery https://www.brickdiscovery.com/mcp

ChatGPT

Settings → Connectors → Create → add this URL as a custom connector:

ChatGPT connector URL

https://www.brickdiscovery.com/mcp

Cursor

Add to .cursor/mcp.json:

Cursor — .cursor/mcp.json

{
  "mcpServers": {
    "brickdiscovery": {
      "url": "https://www.brickdiscovery.com/mcp"
    }
  }
}

Generic remote MCP / streamable HTTP

Generic MCP JSON

{
  "mcpServers": {
    "brickdiscovery": {
      "type": "http",
      "url": "https://www.brickdiscovery.com/mcp"
    }
  }
}

Available tools

Tool	What it's for
`search_tools`	Free-text and filtered search over the catalog — by category, tag, platform, pricing, status, or API availability.
`get_tool`	Full details for one tool by slug — description, deep dive, editor's review, pros/cons, socials.
`list_categories`	Every category in the catalog with its published-tool count — the valid values for `search_tools`' `categories` filter.
`list_families`	The curated top-level groupings of categories used to organize the catalog.
`get_weekly_changes`	What was added, updated, or changed status in the catalog for a given week.

Troubleshooting

Double-check the server URL for typos — it must be exactly https://www.brickdiscovery.com/mcp, no trailing slash.
Restart your MCP client (Claude Desktop, Claude Code, ChatGPT, Cursor) after editing its config — most clients only read connector config on startup.
If your client shows a connection error, confirm it supports the streamable-HTTP MCP transport — this server doesn't speak stdio or SSE.
Some clients cache the tool list — reconnect or refresh the connector if a tool you expect doesn't show up.

Notes

Tool links returned by the server are /go/<slug> redirects through Brick Discovery, not raw URLs — so clicks from your agent's answers are attributed here.
catalog prose returned by get_tool is partly drafted with ai. A human approves publication, and editorial review can continue afterward.
Looking for a plain JSON API or RSS feed instead? See the developers page.