Skip to main content

MCP Server

Sensemaker provides a Model Context Protocol (MCP) server that lets AI assistants interact with your library.

What MCP Enables​

With the MCP server connected, AI tools like Claude can:

  • Save URLs to your Sensemaker library
  • Search your library for saved content
  • Read article content and highlights from your library

This means you can say things like "save this article to my Sensemaker" or "what did I save about API design?" directly in your AI assistant.

Setup​

Prerequisites​

  • A Sensemaker account at sensemaker.app
  • The MCP server built locally (packages/mcp-server/dist/index.js)
  • An AI tool that supports MCP (Claude Desktop or Claude Code)

Claude Desktop​

Add the following to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
"mcpServers": {
"sensemaker": {
"command": "node",
"args": ["/path/to/sensemaker/packages/mcp-server/dist/index.js"],
"env": {
"SENSEMAKER_EMAIL": "[email protected]",
"SENSEMAKER_PASSWORD": "your-password"
}
}
}
}

Replace /path/to/sensemaker with the actual path to your local clone of the repository.

Claude Code​

Add the following to .claude/settings.json in your project root:

{
"mcpServers": {
"sensemaker": {
"command": "node",
"args": ["/path/to/sensemaker/packages/mcp-server/dist/index.js"],
"env": {
"SENSEMAKER_EMAIL": "[email protected]",
"SENSEMAKER_PASSWORD": "your-password"
}
}
}
}

Environment Variables​

VariableRequiredDescription
SENSEMAKER_EMAILYesYour Sensemaker account email
SENSEMAKER_PASSWORDYesYour Sensemaker account password
SENSEMAKER_API_URLNoAPI base URL (default: https://api.sensemaker.app)

Available Tools​

ToolDescription
search_librarySearch your library by query string. Returns matching items with titles, URLs, and metadata.
get_itemRetrieve the full content of a saved library item by ID.
get_highlightsRetrieve highlights for a specific item or across your entire library.
save_urlSave a URL to your Sensemaker library. Content is extracted and processed automatically.

Privacy​

MCP operations run with your credentials and can only access your own library. Content retrieved via MCP is passed to the AI assistant you're using — the same privacy considerations as reading the content yourself.

Source Tracking​

Items saved via MCP are tracked with source: api in your library, so you can filter for content saved through AI assistants vs. manual saves.