openworker vs rowboat
Andrew Ng's local-first desktop AI coworker: give it an outcome and it works across your files, terminal and 25+ apps — Slack, Jira, Notion, Gmail — then hands back the finished deliverable. — versus — Desktop AI coworker (YC S24) that indexes email, meetings and Slack into a living backlinked knowledge graph, then acts on it — email client, browser, meeting notes, background agents, code mode.
Both are desktop AI coworkers that act on your real accounts. Rowboat indexes email, meetings and Slack into a backlinked knowledge graph first; OpenWorker skips the index and drives your apps and terminal directly, gating every write behind approval.
| openworker | rowboat | |
|---|---|---|
| Stars | 15k | 17k |
| Forks | 2.1k | 1.7k |
| Language | Python | TypeScript |
| License | MIT | Apache-2.0 |
| Last activity | yesterday | 3 days ago |
| Topics | agents | agents, memory |
| Curated connections | 4 | 4 |
openworker — the curator's take
Use it when you want an agent that produces artifacts rather than chat: it plans, works across your desktop and connected apps, and stops for approval before sending, writing or running anything. The agent loop, conversations, connector tokens and model keys all stay in the app's local store; bring your own key for 13-plus providers or point it at Ollama. Scheduled automations park their questions in an inbox instead of acting alone. Caveats: it's an open beta desktop app (macOS signed and notarized, Windows builds not yet code-signed), a small cloud service brokers connector OAuth unless you paste credentials by hand, and 'finished work' is only as good as the model you plug in.
rowboat — the curator's take
The thesis is memory that compounds vs retrieval that starts cold: everything you touch gets indexed into an Obsidian-style backlinked graph stored as plain local Markdown — inspectable, editable, yours — and the work surfaces (email client that drafts replies with full context, isolated browser, local meeting transcriber, event/schedule-triggered background agents) all read from it. Code mode drives Claude Code or Codex with that context. BYO model including Ollama. Use it if you want one desktop app to BE the AI layer over your work life. When NOT: it's maximalist by nature — email + browser + meetings + code in one young app means breadth outruns polish in places, and Google-service setup is a manual OAuth dance. Data stays local; that part they got exactly right.