Automatic Memory Extraction
Your agents retain important facts from every conversation, with no manual intervention.
Every interaction with an agent generates valuable information: user preferences, decisions made, project context, technical constraints. Until now, this information was lost at the end of each session.
The Extract-Memories plugin automatically analyzes the content of every completed session and extracts significant facts. These facts are structured, deduplicated, and stored in the agent's persistent memory. In subsequent sessions, the agent has this context available without the user needing to re-explain everything.
Extraction is selective: only information with lasting value is retained. Ephemeral details (intermediate phrasings, abandoned attempts) are discarded to maintain a clean, usable memory.
Session ends
The plugin triggers automatically when a conversation ends or reaches a length threshold.
Content analysis
A specialized model identifies significant facts, preferences, decisions, and context elements.
Structuring
Each fact is categorized (preference, decision, constraint, contact, etc.) and associated with metadata (date, source, confidence).
Deduplication
Already-known facts are merged or updated rather than duplicated.
Persistent storage
Validated memories join the long-term store, available for all future sessions.
Personal assistant
A user mentions they prefer short, structured responses. The agent remembers and adapts its style in all subsequent sessions.
Project tracking
Technical decisions made across sessions (framework choices, deployment constraints) are automatically consolidated. No context loss between sprints.
Customer support
The history of issues raised and resolved for a client is retained. The agent can reference past incidents without the client needing to re-explain them.
featurePages.extractMemories.config
- Available for all EasyClaw v2 agents
- Active by default on Max, Eva, and Neo
- Requires a memory backend (mem0 or Cognee)