Rowboat: An Open-Source AI Coworker That Keeps Your Data on Your Machine
The trending open-source desktop app indexes your email, meetings, and notes into a local knowledge graph — and works with local models via Ollama or LM Studio.
An open-source project called Rowboat is drawing attention on Hacker News as a local-first alternative to AI desktop assistants like Claude Desktop. The pitch: a desktop AI coworker that builds a memory of your work — and stores all of it on your own machine.
Here’s how it works. Rowboat indexes your email, meeting transcripts, Slack messages, and AI conversations into a knowledge graph — think of it as an automatically maintained web of linked notes, in the style of the note-taking app Obsidian, where related topics, people, and projects point to each other. Everything is stored locally as plain Markdown text files you can open, edit, or delete with any editor. On top of that memory sit work surfaces: an email client, notes, a browser, a code mode, and a meeting note-taker that produces live transcripts and summaries.
The part that will interest privacy-minded readers most: Rowboat is bring-your-own-model. You can run it entirely with local models through Ollama or LM Studio — free tools that run AI models directly on your computer — or plug in a hosted model with your own API key. Either way, the project says, your data stays in your local Markdown vault. The app is available for Mac, Windows, and Linux, and the code is public on GitHub.
What’s behind the buzz? There’s growing unease about how much personal context the big AI assistants accumulate — and the fact that this context usually lives on someone else’s servers. Rowboat’s answer is the “local-first” pattern: the intelligence can come from the cloud if you choose, but the memory — the genuinely sensitive part — stays yours, in files you control. It’s the same philosophy driving tools like Obsidian and the local-AI movement generally, now applied to the agent category.
Worth knowing before you get excited: this is a young open-source project from a small team (a Y Combinator startup), not a polished consumer product. Expect rough edges, and remember that fully local models are noticeably less capable than frontier cloud models — the usual trade-off between privacy and horsepower.
What this means for you: If you’re curious about AI assistants but hesitant to hand your email and meetings to a cloud service, Rowboat is one of the more interesting experiments to watch — and trying it costs nothing. For power users, the plain-Markdown storage is the killer detail: no lock-in, your data outlives the app. For everyone else, it’s a preview of a choice you’ll likely face more often: convenience in the cloud, or control on your machine.
Sources
Amazon Is Quietly Closing the Original 'Human-Powered AI' Marketplace
Amazon will stop accepting new customers for Mechanical Turk on July 30, 2026 — the crowdsourcing platform that helped train a generation of AI, as generative AI now does that work instead.
