Memory

Explore and manage what your agents remember.

The Memory tab lets you explore the memory system — what your agents have stored, recalled, and learned across interactions.

How agent memory works

Agents automatically store important information from conversations and tasks:

• Key facts and decisions from chat interactions

• Results from completed tasks

• User preferences and corrections

• Cross-agent shared knowledge

Memory is persisted across conversations, so agents build context over time.

Browsing memories

The memory explorer shows:

• Memory entries — individual facts or learnings

• Source — which agent stored it

• Timestamp — when it was created

• Relevance — how often it's been recalled

Searching

Search across all stored memories by keyword or topic. The search uses vector similarity, so natural language queries work well:

• "What do we know about the authentication system?"

• "Previous security findings"

• "Customer preferences"

Managing memories

• Delete — remove incorrect or outdated memories

• Edit — correct factual errors in stored memories

• Pin — mark important memories to prevent pruning

Deleting a memory is permanent. The agent will no longer recall that information.

Memory per agent

Each agent has its own memory scope, plus access to workspace-level shared memories. This means:

• Sentinel's security findings don't clutter Code Reviewer's memory

• Shared knowledge (project context, team preferences) is available to all agents