Frequently asked questions

Memorandai is a desktop Knowledge Studio — a local-first application where you bring your AI conversations and reference material into one workspace, build a layered memory system over time, and work with multiple AI models on a visual Flows canvas. It inverts the typical AI relationship: instead of you visiting a different AI application with bits of context each time, your knowledge stays on your machine and AI integrates directly with it.

No. SaaS usually means three things: the vendor hosts the software, you reach it through a browser, and you pay a recurring subscription while your data sits on their servers. Memorandai is the opposite on all three — it's a desktop app that runs on your own machine, it's a one-time purchase with a perpetual license (no subscription), and your knowledge lives on your own filesystem. The closest label is old-fashioned packaged software: buy it once, own it. The only ongoing cost is the AI model usage you choose, and that's paid directly to the providers through your own API keys — never to us.

ChatGPT keeps your conversations on its servers; Notion AI does the same with its own. Memorandai keeps everything on your filesystem. You bring your own API keys to whichever AI providers you want to use, and your queries go directly from your machine to those providers — Memorandai never sees them or proxies them. The result: AI works with your accumulated knowledge directly, and your knowledge isn't trapped in someone else's product.

No — Memorandai is more of a complement than a competitor. It doesn't build models or try to replace the labs that do; it's an interface and a home for your context that sits on top of whichever models you choose. There's one subtle overlap: because you chat with GPT, Claude, Gemini, or Grok directly inside Memorandai using your own API keys, it can stand in for a monthly ChatGPT or Claude subscription — you're still paying those providers, just pay-as-you-go through their API, and your history lives on your own device instead of tied to your account on theirs.

The real difference is ownership. The major labs build the models and wrap them in their own interfaces, on their own servers. Memorandai is a wrapper that folds up in your pocket, not theirs — a place to work with those same models while building a centralized knowledge base that's yours. You're still working with their core product, the models; you're just doing it on your own terms.

Every conversation, memory, document, and keystone is a file on your filesystem. There's no cloud sync, no Memorandai-controlled server holding your data, no third-party storage. When you route a query to a cloud AI provider, the request goes directly from your machine to theirs — Memorandai is never in the path. You can browse, back up, port, or delete any of it using standard OS tools.

API keys are recommended for the full experience. With keys from OpenAI, Anthropic, Google, and/or xAI, you get cloud chat models, image generation, background processing (consolidation memos, identity work), and Roundtable synthesis. Without cloud keys, you can still run local GGUF models for chat — the app's built-in local inference engine handles that — but with a more limited feature set. Background tasks and some specialty features need cloud providers.

Yes — Memorandai has a built-in local inference engine. Download and run open GGUF models on your own hardware in a few clicks, GPU-accelerated when a compatible GPU is available and CPU otherwise. Local models power chat with no API keys and no per-token cost, fully offline after the initial model download. The trade-off is capability: they cover a more limited feature set than the frontier cloud models, and some background and specialty features still route to cloud providers. For the full breakdown of what runs locally, see Local & Privacy.

Local-only chat works fully offline (after the initial model download). Cloud sessions and background features (Dream Cycle, Daily Insight, keystone mining) require Internet and configured API keys. License activation happens once online — after that, the license check itself works offline.

Memorandai currently runs on Windows 11 and macOS 11+ (Apple Silicon or Intel). Linux isn't supported yet. The current focus is on getting the Windows + macOS experience right for the Founding Member alpha; Linux support may come later if there's demand.

MCP (Model Context Protocol) is an open standard that lets AI applications connect to external tools and data through one consistent interface — a universal adapter between a model and the things it can act on.

Memorandai puts that standard to work in both directions, and lets you build your own. It runs its own MCP server, exposing your knowledge base and the app's capabilities — search, memory, Flows, project data — as tools an outside client like Claude Desktop, Claude Code, or Cursor can call, so another AI application can work with your Memorandai knowledge directly, with your permission. It also connects to other MCP servers, so the models you work with inside Memorandai can reach external tools — a filesystem, a web-search or SEO API, anything MCP-compatible — added in a few clicks over a local command or a hosted remote connection. And when the integration you want doesn't exist yet, the built-in MCP Server Builder drafts a custom one for you with AI assistance. Through all of it your data stays local; MCP just gives trusted tools and clients a structured, permissioned door to it.

Yes — Memorandai Go is the free iOS companion, available on the App Store. It brings Helios to your phone on your own API keys, captures notes, to-dos, and ideas into your desktop archive, and includes an AI Book Club and a small arcade of memory games. It pairs with the desktop app over your own Wi-Fi — end-to-end encrypted, no account, no cloud — and also works standalone with no desktop at all.

The phone and your desktop talk directly over your home network. You pair by scanning a QR code shown by the desktop app; the encryption key travels only inside that QR, and every sync afterward is end-to-end encrypted between the two devices. Your desktop profile is the source of truth — context flows down to the phone, captures flow up, and the phone's own data (chats, notes, game progress) is backed up into your desktop profile, so a deleted or replaced phone can be restored from your own machine. No Memorandai server ever sits in the middle — none exists.

The app itself is free, with no account, subscription, or in-app purchases — with one important caveat. Chatting with Helios uses your own AI provider API key, billed by that provider at their rates; Memorandai is not in the middle and gets no cut. If you want the app to stay entirely free, simply don't add a key — the tools and games all work, but the AI features are then inaccessible. Unlike the desktop, mobile can't yet run free local models — phones don't have the horsepower.

Not yet — it launched iOS-first. Android may follow if there's demand; tell us at [email protected].

Not by piping one provider's memory feature into another's — that isn't possible from the outside. What Memorandai does is more useful for portability: you import your actual conversation history (your ChatGPT export, Claude conversations, and more) into a repository on your own machine, where it's searchable, connected, and available as context to any model you work with — Claude, GPT, Gemini, Grok, or a local one. So instead of your ChatGPT past being locked inside ChatGPT, it becomes portable knowledge you can carry to any model, on your own terms. See Import & Export for supported formats.

Download the app and use it for 14 days at no cost — no credit card required to try. Test the full feature set with your own API keys (Memorandai never sees them or charges you for AI usage; you pay providers directly). At the end of the trial, you can purchase a Founding Member license to keep using the app, or uninstall — your data stays on your machine either way.

A perpetual license for Memorandai Knowledge Studio (Windows or macOS), 3 device activations managed via the customer portal, access to the Founding Member Discord with direct developer support during the Early Access phase, and free updates for the life of V1. One-time $49 — no subscription, no recurring fees. Price increases at the launch of V1.

Memorandai is planned to leave Early Access with the release of version 1.0.0. At that point the Founding Member discount ends and the license price increases. Early Access is expected to run for several months from the public release in April 2026 (version 0.8.0), with new versions rolling out regularly. Check the Download page for the current version.

We operate no servers that receive your data. All Memorandai data — conversations, memories, keystones, documents — lives on your filesystem. When you route a prompt to a cloud AI provider, it goes directly from your machine to that provider, governed by their terms. We never see your API keys, your queries, or your responses.

Your data stays where it always was: on your filesystem, in standard formats (JSON, Markdown). Uninstalling Memorandai removes the app, not your accumulated knowledge. You can back up, port, or delete any of it using standard OS tools — and Memorandai's Data Management view gives you a curated export path on top of that filesystem-level access.