ClaudeMemory© Documentation

Episodic memory vertical for AI agents and tools.

Purpose

Give agents a stable, queryable memory of past work.

Persistent Episodes

Store important interactions as episodes.

  • Tasks and requests
  • Tools used and approaches
  • Outcome and lessons learned

Mutation Strategies

Experiment with different ways of solving problems.

  • Focus on historically successful patterns
  • Explore alternatives when needed
  • Hybrid strategies mixing both

Calibration

Align confidence estimates with real outcomes.

  • Group episodes into confidence bands
  • Measure observed success rates
  • Use insights to adjust behavior

Episode Model

State

User request and context.

  • Problem description
  • Project and environment

Action

What the agent did.

  • Type of response
  • Tools and methods used

Outcome

How well it worked.

  • Success flag
  • Short summary of results
  • Key lesson to remember

Try ClaudeMemory© Agent

Search your agent episodic memory shard using natural language.

Results from ClaudeMemory© will appear here once the agent backend is connected.