Rowboat is an open-source, local-first desktop app that connects to your email and meetings, builds a persistent knowledge graph of your work, and uses that context to help you prep meetings, draft emails, and generate documents.

Yes. Rowboat is 100% open-source under Apache 2.0. You only pay for LLM API usage if you choose a hosted model — or use a local model via Ollama for free.

How is Rowboat different from Claude Cowork?

Unlike Claude Cowork, Rowboat runs entirely locally. Your data never leaves your machine, and your knowledge graph is stored as plain Markdown files in an Obsidian-compatible vault.

What LLMs does Rowboat support?

Rowboat works with any LLM — local models via Ollama or LM Studio, or hosted models like Claude, GPT, or Gemini with your own API key.

Does Rowboat send data to the cloud?

No. Rowboat is local-first by design. All data is stored as plain Markdown files on your machine. No proprietary formats or cloud lock-in.

Rowboat: The Open-Source AI Coworker That Builds a Living Knowledge Graph of Your Work

By Prahlad Menon Published 2026-04-09 4 min read

Most AI assistants have amnesia. You explain context, get a response, and next session you start from zero. Rowboat takes a fundamentally different approach: it builds a living knowledge graph from your actual work — emails, meetings, notes — and that knowledge compounds over time.

What Is Rowboat?

Rowboat is an open-source desktop app (Mac, Windows, Linux) that connects to your email, calendar, and meeting notes, then automatically constructs a knowledge graph of people, projects, decisions, and commitments. Instead of reconstructing context on every query, Rowboat maintains it.

Think of it as an AI coworker that actually remembers what happened in last Tuesday’s meeting — without you having to paste a transcript.

Why This Matters (and Why It’s Being Compared to Claude Cowork)

Anthropic’s Claude Cowork made waves as a persistent AI work companion. But Rowboat addresses the two biggest concerns people have with that approach:

Everything runs locally. Your data never leaves your machine. No cloud dependency, no subscription lock-in.
Your knowledge is portable. The entire knowledge graph is stored as an Obsidian-compatible Markdown vault — plain .md files with backlinks that you can browse, search, edit, and back up with any tool you already use.

That second point is the real differentiator. Your accumulated work context isn’t trapped in a proprietary database. It’s files on your disk.

Key Features

Knowledge Graph That Compounds Rowboat doesn’t just index — it builds relationships. People connected to projects, decisions linked to meetings, commitments tracked across threads. The more you use it, the more useful it becomes.

Practical Actions, Not Just Chat

“Prep me for my meeting with Alex” → pulls past decisions, open questions, and relevant threads into a brief
“Build me a deck about our Q3 roadmap” → generates a PDF using accumulated context
“Draft a reply to Sarah’s email” → grounds the draft in your actual history and commitments

Voice-Enabled Record voice memos that automatically get processed into key takeaways and added to the knowledge graph. Voice input and output supported via Deepgram and ElevenLabs.

Live Notes Create notes tagged with @rowboat that stay updated automatically. Track a competitor across the web, monitor a deal across your communications, keep a running summary of any topic.

Any LLM Works with local models via Ollama or LM Studio, or hosted models with your own API key. Swap providers without losing anything — your data stays in the local vault.

MCP Extensibility Connect external tools via Model Context Protocol — web search (Exa), Slack, Linear, Jira, GitHub, Twitter/X, CRMs, databases, and custom internal tools.

How It’s Different from RAG-Based Assistants

Most AI tools use retrieval-augmented generation: they search your documents on demand, pull relevant chunks, and generate a response. This works, but it starts cold every time.

Rowboat maintains long-lived knowledge instead:

Approach	RAG-Based Tools	Rowboat
Context	Reconstructed per query	Persistent, compounding
Relationships	Implicit in embeddings	Explicit, inspectable graph
User control	Black box retrieval	Editable Markdown files
Data location	Often cloud-hosted	Always local

The knowledge graph means Rowboat can surface connections you didn’t explicitly ask about — “By the way, Alex mentioned this same concern in your March call.”

Getting Started

Download from rowboatlabs.com/downloads (Mac/Windows/Linux)
Connect Google services (Gmail, Calendar, Drive) via OAuth
Configure your preferred LLM (local via Ollama, or any API provider)
Optionally add voice (Deepgram), web search (Exa), and external tools (MCP)

Everything lives in ~/.rowboat/ — config files, your Markdown vault, all of it inspectable.

The Bigger Picture

Rowboat represents a growing trend: local-first AI that builds persistent context. We’ve seen this with tools like Lore for personal memory and Accomplish for desktop automation. But Rowboat’s combination of knowledge graphs, Obsidian compatibility, and MCP extensibility puts it in a category of its own.

The Apache 2.0 license means no usage restrictions for personal or commercial use. Your knowledge graph is yours — not a feature you’re renting.

GitHub: github.com/rowboatlabs/rowboat Website: rowboatlabs.com