Meet Rowboat: The Local-First AI Workspace You Need

I just stumbled upon Rowboat, and honestly, it feels like the logical evolution of how we should be using LLMs. We’ve all been using the Claude desktop app, which is great, but it still feels like a glorified chat window. It’s a place you go to ask things, rather than a place where you actually do things.

Rowboat is trying to change that by turning the AI from a chatbot into a full-fledged work environment. Instead of just chatting, you get these "work surfaces." It has a built-in email client that drafts replies in your own voice, a local meeting notetaker (think Granola but more integrated), and even a dedicated browser instance. It’s essentially trying to be your entire OS for productivity.

What really caught my eye is the "parallel coding" feature. You can spin up multiple instances of Claude Code or Codex within the app, and it uses an Agent Client Protocol to orchestrate them. If you're a dev, this sounds like a massive time-saver. Plus, it has an Obsidian-style note-taking system that uses a knowledge graph to keep all your context organized.

The best part for privacy nerds like me? It’s local-first. Everything is stored as plain Markdown files on your own machine. It’s open-source (Apache-2.0) and gives you the freedom to plug in any LLM—you can even run local models via Ollama or LM Studio if you don't want to rely on the cloud.

The real "killer feature" seems to be the ability to build your own custom work surfaces. You can essentially create web apps within Rowboat that have their own background agents. Imagine an app that automatically pulls feature requests from your emails and Slack, ranks them, and then uses Claude to draft the initial code based on your existing project notes. That sounds like magic.

Has anyone else tried it yet? I'm curious to see how well the local knowledge graph actually holds up with large datasets.

Check out their repo here: https://github.com/rowboatlabs/rowboat