Takt
Takt is a personal knowledge library that automatically ingests content from YouTube, podcasts, articles, PDFs, and voice notes into one searchable format. It transcribes audio and video, generates AI summaries, and enables semantic search across a user's entire saved library. A remote MCP server connects the library to AI clients like Claude, ChatGPT, and Cursor, allowing those assistants to retrieve and cite personal context directly in responses.
- Artificial Intelligence
- Software Only
Funding
Founders
Founder details are not available yet.
Product
Problem
People save articles, podcasts, videos, and notes for later but rarely retrieve them, because the content sits in disconnected apps with no unified search or memory. Finding a specific idea later requires remembering exact keywords or scrolling through months of unorganized saves, so most saved material is effectively lost.
Solution
Takt is a personal knowledge library that automatically ingests content from YouTube, podcasts, X, Reddit, Substack, Spotify, Apple Podcasts, PDFs, articles, and voice notes into one standardized format. It transcribes audio and video, generates AI summaries, and files everything automatically, so users can search by meaning rather than exact wording. The platform lets users jump back to the exact moment in a source, record voice notes that are transcribed on the go, and ask natural-language questions across their entire library. Takt also exposes a remote MCP server that connects to AI clients like Claude, ChatGPT, Cursor, and Codex, allowing those assistants to search the library, pull relevant excerpts, and cite sources directly in responses. The underlying approach treats a saved library as a record of judgment and priors, giving AI models a personalized weighting that generic training data cannot provide.
Target Audience
Primary users are knowledge workers, researchers, writers, and lifelong learners who accumulate large volumes of saved content and need to retrieve and apply it in their work, writing, and decision-making. Secondary users include AI power users who want their language models to work from a personal context library rather than generic training data.
Features
- Ingests and standardizes content from YouTube, Instagram, TikTok, X, Reddit, Substack, Spotify, Apple Podcasts, PDFs, articles, voice notes, and handwritten notes
- Automatic transcription of audio and video with AI-generated summaries that capture key points
- Semantic search that matches by meaning and intent, not just exact keywords
- Voice note recording with automatic transcription and extraction of key ideas
- Ask Takt natural-language querying across the entire saved library
- Remote MCP server (https://mcp.trytakt.com/mcp) that connects to Claude, ChatGPT, Cursor, Codex, Grok, and other MCP-capable clients for context retrieval and citation
- Seven read-only MCP tools including gather_context, search_notes, get_note, list_notes, list_collections, ask_takt, and whoami, secured via OAuth and row-level database permissions