Rowboat Labs' Open-Source AI Coworker Builds a Living Memory Graph From Your Emails
Rowboat is a free, open-source, local-first AI coworker that builds a persistent knowledge graph from your email, meetings, and notes — and acts on it.
Takeaways
- What it is: Rowboat is a free, open-source (Apache-2.0) desktop AI coworker that builds a persistent knowledge graph from your email, meetings, and notes.
- Local-first: All data is stored as plain Markdown in an Obsidian-compatible vault on your machine — nothing goes to Rowboat's servers.
- Key features: Built-in email client with auto-drafted replies, live meeting transcription, background agents (event- or schedule-triggered), isolated browser, and parallel code agents via Claude Code or Codex.
- Model-agnostic: Works with local models (Ollama, LM Studio) or any hosted API; MCP support lets you plug in custom tools like Exa, Slack, Linear, and GitHub.
- Traction: 16,600+ GitHub stars, front page of Hacker News twice, and a YC S24 backing — built by the team that sold Agara AI to Coinbase in 2021.
- Cost: Completely free; optional features (voice, web search) require third-party API keys (Deepgram, ElevenLabs, Exa).
Rowboat is a desktop AI coworker that does something most AI tools don't: it remembers. Not in the vague, session-scoped way that chat interfaces do, but persistently , building a living knowledge graph from your emails, meetings, Slack messages, and past conversations, and using that graph every time it acts on your behalf. It's fully open-source (Apache-2.0), runs locally on Mac, Windows, and Linux, and has crossed 16,600 GitHub stars.
The project comes from Rowboat Labs, a Y Combinator S24 startup. The founders previously built a deep-learning product for enterprise support reps at Fortune 100 brands, with models that took live notes, suggested replies, and recommended actions while reps were on calls or handling emails , a company Coinbase acquired in 2021. That background shows in the product: Rowboat is less about chatting with an AI and more about giving it enough context to act without being asked.
The core idea: memory that compounds
Most AI tools reconstruct context on demand , they search your transcripts or documents fresh every session. Rowboat instead indexes email, meetings, Slack, and assistant conversations into a living, backlinked knowledge graph organized around people, projects, organizations, and topics. The result is memory that accumulates over time rather than starting cold on every query.

Background agents convert raw data into linked notes while performing entity resolution , merging every mention of the same person or project into one node , and an agent retrieves the relevant nodes before it acts, instead of re-scanning everything from scratch. The whole thing is stored as plain Markdown files in an Obsidian-compatible vault on your machine. Your data exists only on your PC and is never sent to Rowboat's servers. You can open, check, and edit it directly with a text editor. Even if you uninstall Rowboat, the data remains on your device as Markdown files and can be referenced by other tools such as Obsidian or VSCode.