How It Works: Privacy & Security

Mentrex is designed with a Privacy-First architecture. We understand that your code is your intellectual property, and our system is built to protect it.

The Hybrid Architecture

Mentrex operates using a hybrid model that combines Local Processing with Cloud Intelligence.

1. The Local Agent (Your Machine)

The actual "Agent" lives entirely on your computer inside VS Code or the CLI.

File Analysis: It scans your project structure locally.
Command Execution: When it runs builds or tests, it uses your local terminal.
Zero-Upload: We do not upload your entire codebase to our servers.

2. The Secure Gateway

When you ask a question, the agent packages only the relevant context (e.g., the specific file you are editing and your question).

Encryption: This data is encrypted in transit (TLS 1.3).
Anonymization: PII (Personally Identifiable Information) scrubbers run before data leaves your machine.

3. The AI Brain

The prompt is sent to our enterprise-grade AI models.

State-of-the-Art Logic: The model generates the code or answer.
No Training: We have strict agreements that your data is never used to train the public models.

Data Flow Diagram

Terminal

graph LR
    User[Developer] -->|Talks to| Agent[Local Agent]
    Agent -->|Reads (Local)| Files[Your Codebase]
    Agent -->|Secure Request| Cloud[Mentrex Cloud]
    Cloud -->|Reasoning| AI[LLM Provider]
    AI -->|Code Solution| Cloud
    Cloud -->|Response| Agent
    Agent -->|Writes (Local)| Files

Security FAQ

Where are my API keys stored?
- They are stored in your operating system's secure credential vault (Keychain on macOS, Credential Manager on Windows).
Can you see my code?
- No. We only see the snippets you actively send during a chat session. Use the "Context" indicator to see exactly what is being shared.