Built for trust from day one
"Is it safe to let AI control my computer?" is the most common question we hear. It is a good question - and the answer matters more than marketing.
Fazm was designed around a simple principle: you should always understand and control what the agent is doing on your machine. Every architectural decision - from local screen processing to open-source code to visible on-screen actions - exists to earn your trust through transparency, not promises.
How Fazm keeps you safe
Screen analysis runs locally
Fazm uses Apple's ScreenCaptureKit to process your screen entirely on-device. Raw screen data never leaves your Mac - not even temporarily.
Open source - audit every line
Fazm is MIT-licensed and fully open source on GitHub. You do not have to trust our word - read the code yourself and verify exactly what it does.
Permission-based access
Fazm only works with permissions you explicitly grant through macOS System Settings. Accessibility, Screen Recording, and Microphone access are all under your control.
No background data collection
There is no hidden telemetry, no analytics that run without your knowledge, and no data harvesting. If you opt out of analytics, zero data leaves your machine.
Full real-time visibility
You see exactly what Fazm is doing at every moment. It moves your mouse, types on your keyboard, and clicks buttons visibly on screen. Nothing happens behind the scenes.
What Fazm can and cannot do
What Fazm can do
- +Control your mouse and keyboard to interact with any app
- +Read what is on your screen to understand context
- +Access files and folders you explicitly grant access to
- +Fill out forms, send emails, and navigate websites
- +Remember your preferences and contacts over time
What Fazm cannot do
- ×Run without your explicit permission through macOS settings
- ×Access apps without macOS granting accessibility access
- ×Send your screen content, files, or passwords to the cloud
- ×Operate in the background without visible on-screen actions
- ×Bypass macOS security or access Keychain data
- ×Continue running if you revoke its permissions
Privacy by design
Fazm follows a local-first architecture. Your screen content is processed on your Mac using Apple's ScreenCaptureKit - it never touches a remote server. Your personal knowledge graph, file index, and browsing context all live on your machine.
The only data that leaves your Mac is the intent - a short text description of what you want to accomplish, sent to an AI model (like Claude or GPT) for action planning. Your actual screen pixels, file contents, and personal data stay local.
There are no user accounts required. Telemetry is opt-in only. If you choose not to share analytics, absolutely zero data is transmitted. Your files, your memory graph, your workflows - they all stay on your machine, period.
Stays on your Mac
- Screen capture data
- Personal knowledge graph
- File contents and index
- Browsing context
- Saved workflows
Sent to AI model (text only)
- Your spoken or typed command
- Action plan request
- No screen pixels
- No file contents
- No personal identifiers
Open source transparency
Fazm is fully open source under the MIT license. This is not a marketing decision - it is a trust decision. When an app controls your mouse and keyboard, you deserve to know exactly what the code does.
Every commit, every change, every line of code is publicly available. Security researchers, developers, and curious users can audit the codebase at any time. If there were hidden telemetry or data exfiltration, the community would find it.
Frequently asked questions
Can Fazm access my passwords?
No. Fazm does not have access to your macOS Keychain or any stored passwords. It interacts with apps through accessibility APIs and screen analysis - the same way you do, by looking at what is on screen and clicking. It cannot read password fields that macOS masks, and it never attempts to extract credentials from your system.
Does Fazm send my screen to the cloud?
No. Screen capture and analysis happen entirely on your Mac using Apple's ScreenCaptureKit framework. The raw screen data never leaves your machine. The only thing sent to an AI model is a text description of your intent - for example, 'open Safari and search for flights to Tokyo.' Your actual screen content stays local.
Can I see what Fazm is doing?
Yes, always. Fazm operates visibly on your screen - it moves your mouse, types on your keyboard, and clicks buttons in real time. There are no hidden background processes making changes you cannot see. You can watch every action as it happens and stop it instantly with a keyboard shortcut.
What permissions does Fazm need?
Fazm requires macOS Accessibility permission (to control mouse and keyboard), Screen Recording permission (to analyze what is on screen), and optionally Microphone permission (for voice input). These are all standard macOS permissions that you explicitly grant in System Settings. You can revoke any of them at any time.
Is Fazm open source?
Yes. Fazm is fully open source under the MIT license. Every line of code is available on GitHub at github.com/m13v/fazm. You can audit exactly what the app does, how it handles your data, and verify that there is no hidden telemetry or data collection.
Ready to try Fazm?
Open source, local-first, and built on macOS security primitives. Download Fazm and see for yourself.