
Rowboat
Open-source desktop AI coworker with a living knowledge graph, built-in email, browser, meeting notes, and code mode.
Rowboat is a local-first desktop AI coworker that indexes your work into a living Obsidian-style backlinked knowledge graph. It includes built-in work surfaces for collaborating with AI: an email client that auto-drafts responses using work context, a built-in browser isolated from your main browser, a meeting note-taker with live transcription, and a code mode for parallel coding agents.
Unlike most AI tools that reconstruct context on demand, Rowboat maintains long-lived knowledge that compounds over time. Context accumulates, relationships are explicit and inspectable, notes are editable by you, and everything lives on your machine as plain Markdown. It supports bring-your-own-model (Ollama, LM Studio, or any API), MCP tool integrations, and background agents that run on events or schedules.
Build with:
Frequently Asked Questions
Is Rowboat a notes app, or something else?
Rowboat is a desktop AI coworker with memory — closer to an agentic operating surface than a wiki. It's Apache-2.0 licensed, local-first, and built with TypeScript, Python, and a Swift desktop shell. Beyond notes, it bundles an email client, a meeting note-taker, a built-in isolated browser, a "Code Mode" for Claude Code/Codex, and a living knowledge graph.
How does Rowboat's "knowledge graph memory" actually work?
Rowboat distills email, meeting transcripts, Slack messages, and assistant chats into plain-Markdown notes that auto-link to each other, Obsidian-style. People, companies, projects, and topics become nodes in a live knowledge graph that updates as you work. Crucially, the notes are files on your disk — readable, portable, and editable by you rather than locked inside a model.
Can I run Rowboat offline and bring my own AI model?
Yes — Rowboat is local-first by design: everything runs on your machine and your data never leaves it. It supports local models via Ollama or LM Studio as well as hosted models with your own API key, and you can swap models without losing data. Optional integrations like Deepgram (voice input) and ElevenLabs (voice output) are configured via simple JSON files in ~/.rowboat/config/.
What does Rowboat do that a normal notes app doesn't?
It includes a local meeting note-taker that captures mic and speaker audio directly (no bot joins the call) and produces a live transcript plus a Markdown summary. Background agents fire on events (new email, meeting ended) or schedules and can search the web, drive the built-in browser, and write code via Claude Code or Codex. You can also describe a personal app — a CRM, a runway tracker — and Rowboat's copilot will build and maintain a small web app fed by your data and agents.
Looking for contributors
This project is actively seeking help, join the community!
OpenAltFinder Score
Project Details
Repository details
Updated 10/11/2026, 1:00:43 AM
View RepositoryRepository activity
Compare Rowboat with
Similar open source alternatives
Jotted
Self-hosted tasks and notes management with offline support, push reminders, and full-text search in a single Docker container.

Tachyon Transcripts
100% local Windows meeting transcription tool with real-time captions, speaker diarization, and offline Whisper processing.
Otter.ai
Mnemo
Mnemo is a free, open-source desktop study app combining block-based notes, FSRS flashcards, and mind maps with no accounts or tracking.
Quizlet
NoteDiscovery
A self-hosted knowledge base for note-taking with markdown support, Zettelkasten methodology, and Docker deployment.

HedgeDoc
Real-time collaborative markdown note editor you can self-host — write together with rich formatting and seamless sharing.
Google Docs
Many Notes
A Markdown note-taking web application designed for simplicity with vaults, collaboration, and PWA support.





