Personal MCP

Example 1

Get to know the user

User prompt:

“What do you know about me from Personal Hub?”

What happens:

1. Claude calls get_profile_summary
2. Receives the user's name, location, languages, shared dimensions, and a summary of available data
3. Responds with a personalized greeting using the user's actual profile data

Example tool response:

{
  "user": "Marius",
  "location": "Oslo, Norway",
  "languages": ["Norwegian", "English"],
  "shared_dimensions": {
    "basics": { "filled_fields": 4, "total_fields": 7 },
    "interests": { "count": 12 },
    "preferences": { "filled_fields": 2, "total_fields": 4 }
  },
  "data_version": "2026-03-19T14:32:00Z"
}

Example 2

Use interests for recommendations

User prompt:

“Based on my interests, what podcasts might I enjoy?”

What happens:

1. Claude calls get_dimension("interests")
2. Receives the user's full interest list
3. Generates podcast recommendations tailored to the user's actual interests, not generic suggestions

Example tool response:

["Technology", "Data Privacy", "Music Production",
 "Sailing", "Nordic Mythology", "Philosophy",
 "Entrepreneurship", "AI Ethics", "Cooking",
 "Board Games", "Hiking", "Architecture"]

Example 3

Reference past conversations

User prompt:

“What did I discuss with ChatGPT about learning Norwegian?”

What happens:

1. Claude calls get_chat_context("learning Norwegian")
2. Receives relevant excerpts from the user's uploaded ChatGPT conversation history
3. Summarizes the past discussion and can build on it

Example tool response:

{
  "query": "learning Norwegian",
  "matches": [
    {
      "source": "chatgpt",
      "content": "user: What are the best apps for learning Norwegian?\nassistant: For Norwegian specifically, I'd recommend...",
      "metadata": {
        "conversationTitle": "Language Learning",
        "date": "2025-08-15"
      }
    }
  ]
}

Example 4

Efficient context updates

User prompt:

“I updated my preferences yesterday. What's new?”

What happens:

1. Claude calls get_recent_changes with the timestamp from a previous conversation
2. Receives only the changed data (e.g., updated preferences)
3. Responds with awareness of the specific changes without re-reading all data

Example tool response:

{
  "changes_since": "2026-03-18T10:00:00Z",
  "updated_dimensions": [
    {
      "dimension": "preferences",
      "data": { "communicationStyle": "Direct and concise" },
      "updated_at": "2026-03-19T09:15:00Z"
    }
  ],
  "data_version": "2026-03-19T14:32:00Z"
}

Example 5

Search across all shared data

User prompt:

“Do you know anything about my dietary preferences?”

What happens:

1. Claude calls search_context("dietary preferences")
2. Searches across all shared dimensions for relevant matches
3. Returns any matching data from preferences, health, or other dimensions that mention dietary information

Example tool response:

{
  "query": "dietary preferences",
  "results": [
    {
      "dimension": "preferences",
      "data": {
        "dietaryPreferences": "Pescatarian, avoids dairy"
      }
    }
  ]
}

Example 6

AI learns a preference

User prompt:

“I prefer concise explanations with practical examples. I don't like long theoretical introductions.”

What happens:

1. The AI responds to the conversation normally
2. The AI offers: "Want me to save your communication preferences to your Personal Profile?"
3. If the user agrees, the AI calls suggest_profile_update
4. The suggestion appears in Personal Hub for review
5. The user approves, and the preference is saved to their encrypted vault
6. Every connected AI service now knows this preference

Example 7

Save a conversation summary

User prompt:

“Let's save the key points from our discussion today.”

What happens:

1. The AI calls suggest_note with a title and summary
2. The note appears in Personal Hub for review
3. The user can edit the summary before saving
4. Once approved, the note is encrypted and stored
5. Future conversations can reference this note via search_context