Give your AI assistant your taste with MCP
Ask an assistant for a landing page and you get the average of everything it has read. Ask it to work from the dozen references you saved last month and you get something nearer to what you meant. MCP is how an assistant reads your references.
What MCP is
The Model Context Protocol is a standard way for an AI app to use tools and data outside the chat. A server describes what it can do, such as search a vault or read one save, and the assistant calls it when a request needs it.
What the assistant can do with a vault
Curatively’s server lets an assistant search your saves in your own words, open one with its image, list the latest, browse by format, source or colour, and find saves close to another. If you choose Read and save when you connect, it can also save a link, photo or note for you. It cannot publish or delete anything.
Connect it
The setup page has a button for Claude, steps for ChatGPT and Cursor, and the address for any other MCP app. The short versions:
- Claude: Add to Claude from the connect page. Free plans include one custom connector.
- Claude Code: run the command below, then run
/mcpand choose Authenticate. - Cursor: use Add to Cursor on the connect page, or add the server to
.cursor/mcp.json. - Claude Desktop: use the connector, or the local package described on the MCP page.
claude mcp add --transport http curatively https://curatively.io/api/mcpThe local package, curatively-mcp on npm, runs the same tools on your own computer: npx -y curatively-mcp login, then npx -y curatively-mcp.
Things to ask
Finding a save by its tags, colours or description depends on the catalogue. Plus catalogues everything you save; a free account has a cataloguing allowance. Every MCP tool is free on every plan.
- “Find the green chair I saved in June.”
- “Use my Interiors vault as inspiration for this page.”
- “Build the settings screen in the style of that dashboard I saved.”