Memories API

Access users' LLM conversation data from ChatGPT, Claude, and other AI assistants.

Updated 27 December 2024
memoriesapichatgptclauderaw-databeta

RAW Memories API — Beta Feature

The RAW Memories feature allows developers to request access to users' LLM conversation data from platforms like ChatGPT, Claude, and other AI assistants. This provides rich contextual data about user preferences, communication patterns, and AI interaction history.

Note

Basic profile data (name, email, account details) is always included in all requests.

Current Sources

Platform Data Available
ChatGPT Conversations, ChatGPT apps, soon Browser history
Claude Chats and Memories
Grok Chats and Memories
Gemini Chats and Memories

Coming Soon: Perplexity, Character AI

Usage Modes

Mode 1: RAW Data Only

Request only LLM conversation data + basic profile data. Shows only LLM connections in the UI.

<OnairosButton 
  requestData={['rawMemories']}  // LLM data + basic (always included)
  rawMemoriesOnly={true}         // Show only LLM connections
  webpageName="Your App"
  onComplete={handleRawMemories}
/>

Mode 2: RAW Data + Normal Data Requests

Request LLM conversation data plus additional user insights.

<OnairosButton 
  requestData={['rawMemories', 'personality', 'preferences']}
  rawMemoriesOnly={false}
  rawMemoriesConfig={{
    platforms: ['chatgpt', 'claude', 'gemini'],
    dateRange: {
      from: '2024-01-01',
      to: '2024-12-31'
    },
    includeMetadata: true,
    maxConversations: 100
  }}
  webpageName="Your App"
  onComplete={handleRawMemories}
/>

Parameters

requestData Array

  • 'rawMemories' — Include LLM conversation data
  • 'personality' — Include AI-analyzed personality traits
  • 'preferences' — Include user preferences and interests

rawMemoriesConfig Object (Optional)

{
  platforms: string[],        // ['chatgpt', 'claude', 'gemini', 'perplexity']
  dateRange: {
    from: string,             // ISO date string
    to: string                // ISO date string
  },
  includeMetadata: boolean,   // Include conversation metadata
  maxConversations: number,   // Limit number of conversations
  minLength: number           // Minimum conversation length (messages)
}

Response Format

Success Response

{
  success: true,
  data: {
    basic: {
      name: string,
      email: string,
      profileInfo: object
    },
    rawMemories: {
      conversations: Conversation[],
      metadata: {
        totalConversations: number,
        platforms: string[],
        dateRange: { earliest: string, latest: string },
        processingInfo: {
          filtered: number,
          included: number,
          totalMessages: number
        }
      }
    },
    personality?: object,
    preferences?: object
  }
}

Conversation Data Structure

[
  {
    id: string,              // Unique conversation UUID
    title: string,           // Conversation title
    created: number,         // Unix timestamp
    model: string,           // AI model used (e.g., "gpt-4", "claude-3")
    platform: string,        // "chatgpt" | "claude" | "gemini" | "perplexity"
    messages: Message[]      // Array of all messages
  }
]

Privacy & Security

HIPAA COMPLIANCE

Via Delve — ETA early week of 27th

Data Processing

  • All conversation data is processed locally when possible
  • Sensitive information is automatically filtered
  • Users have full control over what data is shared

Data Retention

  • Raw conversation data is not stored by Onairos
  • Only processed insights and preferences are retained
  • Users can request data deletion at any time

Beta Limitations

  • Maximum 100 conversations per request
  • Processing time may be longer for large datasets
  • Some metadata fields may not be available for all platforms
  • Rate limiting applies to prevent abuse

Feedback

PreviousHuman APINextAPI Reference