Security
Privacy and ownership without compromise.
Keep your knowledge under your control, with the freedom to choose how AI works with it. Start locally, allow cloud processing where it suits the work, and set boundaries for each workspace.
Built for trust
Privacy by default.
Your knowledge should become an asset you own. These are the promises behind the workspace you build.
Nevertrained on.
Brainoid never uses your knowledge to train AI models.
Neveran advertising ID.
No ads, no tracking identifier, no profile of you.
Offlineif you want it.
Local AI runs on your Mac, without cloud processing. After account setup and model installation, you can use it offline.
Frontierwhen you need it.
Choose capable cloud models when your work calls for them. Designated open-model routes offer zero retention and US or EU processing.
Offline use needs a Mac that meets the requirements, installed local models and an existing signed-in session. Cloud models and connected services are optional; their privacy terms depend on the route you choose.
Your security, your rules
Your models. Your choice. Per workspace.
Choose which kinds of AI may receive knowledge from each workspace. Set the processing rules for the work, then choose a model within those boundaries.
Local AI
Model processing stays on your Mac. Choose local-only processing for knowledge you want to keep on the machine.
Zero-retention cloud AI
Open-weight models in the cloud. You set the terms (retention, region, price limit), and a request that cannot meet them fails instead of going elsewhere.
Proprietary cloud AI
Proprietary models, when a workspace allows them. Their providers’ terms apply; this is not a zero-retention route.
Your own subscription
Use a supported AI subscription or provider account you already pay for. Its provider’s terms and available models apply.
Rules that processing must meet
For designated zero-retention routes, choose the available processing region and price limit. If a request cannot meet your requirements, it fails instead of being sent along a different route. Zero retention does not cover every cloud model or connected service.
Change models. Keep your knowledge.
Your workspace’s files and accumulated memory stay with you when you change models. Notes are Markdown files, and memory can be exported. Model choice and ownership of your knowledge work together.
Workspace separation
Separate workspaces. Separate boundaries.
Give each client, project or part of your life its own knowledge, memory and conversations. Each workspace has its own AI Employee configuration and rules; choosing a different Persona does not widen the active workspace’s knowledge boundary.
Client: Northwind
AI Employee: Account manager
- Local AI only
- Telegram off
Own notes · own memory · own chats
Research
AI Employee: Research analyst
- Zero-retention cloud AI
- Web search on
Own notes · own memory · own chats
Personal
AI Employee: Personal assistant
- Local AI only
- Telegram on
Own notes · own memory · own chats
These examples illustrate different workspace policies. Knowledge is not shared across workspaces. Only the active workspace’s AI Employee runs at a time; messaging and employee capabilities require a configured, supporting build.
Controlled access
An AI Employee within boundaries you control.
Let an employee act on your knowledge while keeping its environment, workspace and connections explicit.
Its own isolated computer
The AI Employee works in a separate virtual computer, rather than acting directly on your Mac. Its environment contains the tools and access you configure.
One active workspace
It works with the active workspace’s knowledge. Switching workspaces replaces that boundary; switching Persona does not widen it.
Access you allow
The employee reaches the files, tools and destinations you permit. Connections are part of your setup, rather than general access to your computer.
A configuration you control
Choose its Persona, instructions, model and supported agent engine per workspace. You can see what it is configured to run.
Communication happens in Chat or through the messaging services you connect, including WhatsApp and Telegram. Those services have their own terms. Scheduled work needs a running employee; continuous monitoring needs an always-on deployment.
Agent engine and setup choices
A supporting build lets you choose Hermes Agent or OpenClaw per workspace. Give the employee a Persona with a name, instructions and its own model. Setup and available connections determine what it can do.