Rowboat

Rowboat's argument is that a personal assistant and a team space should be the same product, and the homepage's clearest demonstration is the @rowboat summon: John asks in a shared Space for everything discussed about a feature, and his Rowboat privately searches his Brain, emails, meetings, and notes on his machine before returning an answer to the room. That is the technical crux — private context, shared output — and it is a harder design problem than either half alone. The second strength is that the work and the conversation stay in one place. Spaces hold whiteboards, markdown docs, presentations, and files next to the thread, so a referral flow drafted in conversation can be edited on the same canvas, with background agents running four at a time in the background. This is why Rowboat reads as a coworker rather than a chat window: it acts on schedules and events, not only on prompts, and it can turn a decision into code for review. The honest costs. Rowboat ships as a Linux x64 download, so Mac and Windows teams need a Linux machine or VM — the vendor's own download button says Linux (x64) and nothing else. Model access is bring-your-own through Ollama or LM Studio, which means you wire up and maintain local inference yourself. Context lives on your machine, so backups, disk, and upgrades are yours too. None of that is a defect if you are the intended buyer; all of it is friction if you are not. Where it fits: engineering-led teams who want their email, meetings, and decisions feeding one memory they host, and who have at least one person happy to run Docker and a model endpoint. Where it doesn't: teams that need a browser-only, nothing-to-install experience, or who want their assistant's conversation history managed for them. If a managed memory tool's convenience is worth more to you than local storage of your inbox, Rowboat's core promise will feel like overhead. If it isn't, Rowboat is one of the few tools in its class where the privacy claim is architectural rather than a policy clause.