Skip to main content

​
How It Works

When you ask the agent to remember something, it saves that learning to a project-level file (CLAUDE.md or AGENTS.md) so it persists across tasks. Every agent reads these files at the start of a task. Learnings from previous tasks are available immediately, with no extra configuration needed.

​
What Gets Remembered

You decide what’s worth saving. Typical examples:
  • Build and test commands
  • Framework and language conventions
  • File layout and architecture patterns
  • Past mistakes and what fixed them
Only non-obvious learnings that aren’t already documented are saved.

​
Availability

AgentMemory support
Claude CodeSupported
CodexComing soon
OpenCodeNot supported
We’re also working on making memory portable across agents, so knowledge gained by one CLI carries over when you switch to another.