# Agent Proxy Ultra Secure Level 1

- **Personal Agents (PA):** Linked to each human, manage tasks and access to data.  
- **Worker Agents:** Specialized executors triggered by Personal Agents.  
- **LLMs (Brains):** Provide reasoning and advanced language/knowledge.  
- **Proxy Agents:** Act as intermediaries between Personal Agents.  
  - Changeable identity  
  - Preserve privacy & sovereignty  
  - Prevent direct PA-to-PA exposure  


### Architecture Diagram

```mermaid
%%{init: {'theme':'dark'}}%%
graph TD

    H[Human User] --> PA[Personal Agent]

    PA --> LLM[(LLM / Brain)]
    PA --> WA1[Worker Agent: Research]
    PA --> WA2[Worker Agent: Coding]

    %% Proxy Layer
    PA --> Proxy[Proxy Agent]
    Proxy --> PAX[Other Personal Agent]

    %% Worker examples for other PA
    PAX --> WA3[Worker Agent: Analytics]

```

## Proxy Agent

- **Function:** Acts as a go-between for Personal Agents, ensuring no direct connection.  
- **Identity:** Can change or mask identity to enhance privacy.  
- **Security:** Prevents direct exposure of Personal Agents to each other.  
- **Use Cases:** Ideal for sensitive communications, cross-organization tasks, or when anonymity is needed.

Difference with other implementations

- works for all hero functionality, not just chat
- can be implemented multi level (proxy of proxy)

## Flow Example

- Proxy Agent changes public key (max 1x per week)
- User can chose if personal secure friends (a white list) receive the updated public key automatically