Skip to main content
Dify agents have intelligent memory that helps them learn from every conversation. Memory makes interactions more personal and context-aware over time.

​
Types of Memory

Conversation Memory

What was discussed in the current and recent threads. Agents maintain context within and across conversations.

Entity Memory

People, organizations, and things mentioned in conversations. The agent remembers who “John” is and what you discussed about him.

Fact Memory

Key facts and preferences extracted from conversations. “User prefers reports in table format” or “Project deadline is March 15.”

Workspace Memory

Shared knowledge across all agents in a workspace. Information available to every agent, not just one.

​
How Memory Works

After each conversation, the system automatically:
  1. Extracts important entities, facts, and preferences
  2. Stores them with semantic embeddings for future retrieval
  3. Retrieves relevant memories when they match a new conversation topic
You don’t need to configure anything — memory extraction happens automatically.

​
Memory in Action

First conversation:
You: “I’m working with the Acme Corp team on the Q2 marketing campaign. The budget is $50k.”
Later conversation:
You: “What do you know about my current project?” Agent: “You’re working with Acme Corp on the Q2 marketing campaign with a $50,000 budget.”

​
Managing Memory

  • Memories are visible in the agent’s Memory tab
  • You can delete specific memories that are incorrect
  • Workspace memory is shared — changes affect all agents
  • Memory is automatically updated as new information appears
Be mindful of sensitive information in conversations. Agents will remember and may reference it in future interactions.