Skip to main content

Agent Memory

Agent memory allows agents to retain information across conversations and task executions.

How It Works

Memory entries are key-value pairs stored per workspace and project. Agents can read from memory during conversations to provide context-aware responses.

Managing Memory

Navigate to the Agents page and open the Memory tab:
  • Create — Add a new memory entry with a key, value, and category
  • Edit — Update an existing memory entry’s value
  • Delete — Remove a memory entry

Categories

Organize memory entries by category:
  • general — Default category for miscellaneous facts
  • preferences — User or workspace preferences
  • context — Background information for the agent

Use Cases

  • Store customer preferences that agents reference during support conversations
  • Keep project-specific context that agents use when executing tasks
  • Maintain running summaries of important decisions or outcomes
Memory is scoped to a project. If you switch projects, agents will only access memory entries belonging to the current project.