Start with memory. Upgrade when you are ready to act.

Because the missing piece was never model quality. A frontier model is already general — it is just not general about you. ShogunAI adds the two things that make any model yours: a memory of your actual work, and the ability to act on it. We do not train a smarter model. We give the ones you already use the context and the hands.

Text, where it came from, and when — and, with visual recall on, screenshots of the windows text capture cannot read. ShogunAI reads the accessibility layer macOS exposes to screen readers first; a compressed screenshot is taken only where that yields nothing. Screenshots stay encrypted on your Mac for a retention period you choose — presets from one to seven days, or a bounded custom duration — and are deleted automatically at that age. You can view or delete them any time before then. Meeting audio is never written to disk — only the transcript is kept.

Only the specific slice a request needs — the thread you asked to summarize, the context for a draft — never your raw memory. Each one is written to a traceability log with its purpose, and marked when it passes through a third party. You can read that log whenever you want.

No. ShogunAI can run on the assistant plan you already pay for, inside that plan’s limits. Bringing your own API key is the alternative, not the requirement. Either way you pick the provider, and you can switch it without losing a day of memory.

A recorder captures one meeting. A notes app stores what you were disciplined enough to write down. ShogunAI holds the whole working day without you maintaining it — and then does something with it. That last part is the difference that matters: it finishes the work instead of describing it.

Yes. It is built on native macOS capabilities — the accessibility layer, the notch panel, the system keychain — and runs on Apple Silicon. A version for another platform would be a different product, not a port.