All your AIs.
On your computer.
Nowhere else.

There's a free trial: download the app and try it with your own API keys at no charge. When you're ready to keep it, it's a $250 one-time purchase — pay once, own it, no subscription. AI usage is billed separately by the provider whose API key you use; with local models via Ollama or LM Studio there are no usage costs at all.

Your license key is sent to you by email automatically after purchase, to the address you used at checkout. Enter it in the app under Tools → License to unlock the full version. Not there after a few minutes? Check your spam folder — and if it's still missing, email [email protected] and we'll resend it.

Everything — chats, documents, vector index, API keys — is stored locally on your computer. When you send a message, the request goes directly from your machine to the provider you configured. We never see, collect, store, or train on your content, and telemetry is disabled in our builds.

OpenAI, Anthropic, Google Gemini, Mistral, Groq, OpenRouter, Perplexity, Cohere, Azure OpenAI and more — plus fully local models through Ollama, LM Studio and compatible runtimes. You can set a system-wide default and override the model per workspace.

The web app is a hosted multi-AI workspace: one subscription, premium models included, AI collaboration and Studios. The desktop app runs on your own computer with your own keys — the choice for sensitive documents, offline work, and maximum control. Many users run both.

No — it's a separate product with its own feature set. The desktop app may not include everything you know from the web app, in particular AI Collaboration (AI-Team modes), side-by-side model comparison, the Document, Presentation, Excel and Image Studios, and other web-only features. The desktop app focuses on private, local multi-model chat, document chat and agents; for the full collaboration feature set, use the web app.

Three ways of working. Chat mode blends your documents with the model's general knowledge — best for brainstorming. Query mode answers strictly from your uploaded documents and tells you when they don't contain the answer — best for factual lookups. Agent mode (@agent) lets the model use tools — web search and scraping, document search, charts, saving files — chosen automatically. All of it with cloud providers or fully local models, stored on your machine.

The honest limits: document chat retrieves the most relevant passages of your files (RAG) rather than feeding whole documents to the model — so "summarize this entire 500-page book in one go" won't work unless it fits the model's context window. Document answers depend on wording and the adjustable similarity threshold. It's single-user — no team accounts, shared workspaces or admin console. It ships no AI of its own — you need an API key or a local model. And agent reliability depends on how well your chosen model handles tool calling.

Only if the document fits your model's context window. Otherwise document chat retrieves the passages most relevant to your question instead of re-reading the whole file. For long documents: ask specific questions, summarize section by section with the agent, or pin a smaller document to force it fully into context.

Yes — fully offline with local models via Ollama or LM Studio: chat, document indexing and document Q&A all run on your machine. Cloud providers and agent web-search naturally need an internet connection.

The desktop app is single-user by design — one installation, one person's workspaces, one computer. No shared logins, roles or admin features. For shared workspaces, centralized billing and seat management, use MultipleChat for Teams on the web.

A 64-bit Mac (Apple Silicon or Intel, macOS 12+) or a Windows 10/11 PC. With cloud providers the app is very light; if you want to run local models, 16 GB RAM or more is recommended.