Trove Documentation

Trove is a personal knowledge base that makes your saved content searchable by AI. It connects to data sources (RSS feeds, Hacker News, Readwise, newsletters, bookmarks), indexes content with vector embeddings, and exposes it through MCP tools and a GraphQL API.

How It Works

Trove has three parts:

• Mac app — Runs connectors locally on your machine to pull in content from your reading sources. Enable the sources you care about, and the app syncs them into your Trove account. Get started with the Mac app.
• Web app — Browse your documents, connectors, and sync status at app.ontrove.sh. A read-only view of everything in your account.
• MCP server — Gives AI assistants (Claude Code, Claude Desktop, Cursor) direct access to your knowledge base. Search, browse, and save content without leaving your editor.

Get Started

Install the Mac App Download the Mac app, enable connectors, and start syncing your reading sources.

Connect an AI Assistant Add Trove to Claude Code, Claude Desktop, or Cursor to search your knowledge base with natural language.

Build an App Query Trove’s GraphQL API directly. Search documents, save content, and manage connectors programmatically.

Features

• Semantic search across all sources. Find content by meaning, not keywords. Vector embeddings surface relevant results regardless of exact wording.
• 7 MCP tools for Claude and Cursor. Search, browse, discover, and save content without leaving your AI assistant.
• GraphQL API. Build integrations, dashboards, or workflows on top of your knowledge base.
• Per-user data isolation. Every user gets their own database. No shared tables, no user_id columns.
• Connectors for RSS feeds, Hacker News, Readwise, and more. Browse all available connectors at trove-connectors.

Learn More

Read about how Trove works to understand how documents, connectors, and search fit together.