Launching today
Rowboat
The AI work app that understands how you work
32 followers
The AI work app that understands how you work
32 followers
Most AI tools need five prompts to get it right. Rowboat needs one. It builds a living knowledge graph from your meetings, emails, and notes, so it already knows your work and how you work. Take meeting notes, prep for your day, draft emails, build dashboards, automate browser tasks, and manage projects with context you don’t have to re-explain.
👋 Hey Product Hunt,
I’m Arjun, co-founder of Rowboat.
We built Rowboat because we kept running into a frustrating pattern with AI tools: they don't actually know your work. You have to keep catching them up on your people, your projects, the decisions you've already made. Every conversation starts from zero.
Rowboat is an AI work app that builds a living knowledge base from your meetings, emails, and notes, then uses that context to actually help you get work done.
A few things that make Rowboat different:
Living knowledge base: Rowboat learns about you, the people you interact with, projects, and workflows from the work you’re already doing. No manual tagging or setup.
Meeting notes that compound: it records meetings, extracts action items and decisions, and feeds that context back into your knowledge graph.
Daily prep that knows your world: get a brief that already understands your calendar, inbox, and what matters this week. Auto-drafts emails, creates presentations, or handles anything you'd normally do.
Batteries included: Rowboat can look up or do things on your computer, use a built in browser, do granular web searches and use the library of external product integrations to get work done.
Local-first and open source: Rowboat runs as a desktop app on macOS, Windows, and Linux. Your data is stored on your own machine.
A core belief behind Rowboat: the real measure of an AI product isn’t how clever the model is. It’s how little effort you have to spend before it gets something meaningfully right.
Context shouldn’t be something you type. It should be something your AI already knows.
We'd love to hear your feedback! Two things we’re especially curious about: (1) What’s the last work thing you had to explain to an AI tool? (2) What would you want it to remember without being told?
Congrats on the launch, Arjun.
"Every conversation starts from zero" is the right diagnosis.
The knowledge graph approach makes sense in theory, but the hard part is whether it stays coherent over time or becomes a cluttered mess of stale context that starts causing wrong assumptions instead of right ones. Local-first and open source is a meaningful trust signal for something ingesting your meetings and emails.
To answer your questions: the last thing I had to re-explain was the status of an ongoing project: who's involved, what was decided, and what's still open. What I'd want remembered without being told is relationship context: not just who someone is, but the history of how we've worked together and what's already been agreed.
What's the graceful degradation story when the knowledge graph gets something wrong and confidently acts on it?
@ryanwmcc1 Great question.
A few things we do:
We automatically create nodes only for high confidence people and organizations. The rest come as suggestions that you can create if you think they're important to track. You're also free to create a tracking note anytime for a person or topic.
The graph is fully transparent. Everything Rowboat knows is stored as markdown on your machine. You can read it, edit it, delete it. No hidden memory.
Recency is weighted. If something changed last week, that takes priority over a conversation from a few weeks ago.
The graph is structured - people, projects, decisions all have typed relationships and timestamps. So Rowboat isn't just pattern matching, it's traversing real relationships with temporal awareness. And if Rowboat does get something wrong, everything it references is linked back to the source (the meeting, email, or note it came from), so you can trace why it made that assumption and correct it.
Honestly though, this is an ongoing problem. Right now the answer is transparency, editability, and letting you control what gets tracked. We think a lot about this and it's only going to get better.
Your example about project status and relationship context is exactly what we're building for. Appreciate the thoughtful question.
This is really interesting - I'd also be curious to know whether you've looked at applying this to physical work settings like laboratories? Experimental workflow/planning, analysis etc.? Good luck with the launch!
@buildingoggles Thanks! We haven't explored lab/experimental workflows yet, but the underlying architecture could work for any knowledge-heavy environment where context compounds over time. Would love to hear more about what that workflow looks like for you.
@arjun_maheswaran Cool! I built a simple experiment weekly scheduler app a while back for grad students (having been one myself not so long ago, I know that juggling and interlocking all the different pieces of an experimental workflow can be challenging, I used pen and paper but why not make it easier? :) This seems like a much more sophisticated solution for workflow management, would be interested to see if you end up taking in a science lab application direction as well. Good luck!
@buildingoggles That's a great use case. Thanks for the pointer, we'll explore this more. I spent a lot of my career building ML models and tracking experiments - I can definitely relate to this. We are having a few interns join us, maybe a good time to explore science lab type workflows.