These are the primitive, low-level tools that our context engineering flows are built upon. While our primary jean_memory tool orchestrates these for you, they are directly callable for advanced use cases where you need precise, deterministic control over the memory layer.

​
Primary Tool: jean_memory

The jean_memory tool is the main interface for context-aware interactions. The interface varies based on your integration method:

​
MCP Integration

For MCP clients (Claude, cursor, custom MCP tools): 
{
  "tool_name": "jean_memory",
  "tool_params": {
    "user_message": "Your query here",
    "is_new_conversation": false,
    "depth": 2
  }
}
Depth Levels:
  • 0: No context (generic knowledge questions)
  • 1: Fast search (quick personal facts, 0.5-1s)
  • 2: Balanced synthesis (conversational AI, 3-5s) - Recommended
  • 3: Comprehensive analysis (research tasks, 20-30s)

​
SDK Integration

For SDK users (React, Node.js, Python): 
await jean_memory({
  user_message: "Your query here",
  is_new_conversation: false,
  needs_context: true,
  speed: "balanced" // Options: "fast", "balanced", "autonomous", "comprehensive"
});
Speed Modes:
  • fast: Direct memory search (0.5-1s)
  • balanced: AI synthesis with Gemini 2.5 Flash (3-5s) - Recommended
  • autonomous: Intelligent orchestration with variable latency
  • comprehensive: Deep document analysis (20-30s)
Learn more about speed modes →

​
Document and Memory Tools

  • store_document(title, content): Store large documents, files, or lengthy content. Perfect for saving meeting notes, documentation, or articles for future reference.
  • add_memories(text): Manually stores specific information, bypassing the AI-powered “Smart Triage.”
  • search_memory(query): Performs a quick keyword search through your memories.
  • ask_memory(question): Asks a simple question about your stored memories.
  • list_memories(): Browses through your stored memories.
  • deep_memory_query(query): Performs a complex analysis across all of your memories to synthesize deeper insights.