teak: personal knowledge vault with MCP-powered AI integration across platforms
teak, from Praveenjuge, is a personal knowledge hub designed to collect, organize, and rediscover ideas and inspirations across devices. The app centralizes notes, links, and snippets and exposes that stored information to language models via a server connection for natural-language queries and retrieval. Its focus is personal memory and active recall. Creative professionals, researchers, writers, and power users can use it to keep a searchable vault of project materials and ideas.
What tasks can you actually use it for?
The app focuses on capturing and re-finding creative material: notes, links, snippets, and project fragments. It provides a unified capture system and platform clients so you can save items from browsers and devices. Platform components include a Next.js web app, Expo-based mobile apps, an Electron desktop client, a Wxt-powered browser extension, and a Raycast extension for quick access and searching.
How reliable are model-driven retrievals?
The project includes an MCP server in its API layer so external assistants can query the vault and produce natural-language answers. Because returned text is produced by connected language models, generated replies reflect the model's outputs rather than representing an authoritative transcription of your notes. Treat model responses as draft material for research or writing and verify facts independently before using them in high-stakes work.
Is it practical to run and extend in real projects?
The codebase is organized as a monorepo that bundles web, API, mobile, and desktop components, which supports coordinated development across platforms but assumes familiarity with multi-package repositories. The backend uses Convex for real-time synchronization. The source is MIT-licensed and open-source so teams can self-host and audit the code. The MCP server component is designed to run in environments that support the Model Context Protocol.
Who should adopt it and the practical trade-off
teak is a practical choice for creative professionals who want an auditable, AI-connected personal vault and can manage model integrations. Expect improved retrieval when paired with a capable language model, and plan verification steps for factual outputs. Users who prefer zero-configuration, purely local note-taking may find the setup heavier than necessary; allocate time for server configuration and model checks as part of your workflow.
