---

name: "agent-session-memory" description: "Guides Cognee session memory, typed memory entries, feedback, agent isolation, and agent-memory decorator workflows." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Cognee Agent and Session Memory

Use this sub-skill when the user wants Cognee to remember conversation state, store agent traces, record feedback, isolate memory per agent, or decorate an async agent with memory retrieval and persistence.

Route here for

• QAEntry, TraceEntry, FeedbackEntry, and SkillRunEntry payloads.
• recall(..., session_id=..., scope=...) behavior.
• @cognee.agent_memory(...) decorator configuration.
• cognee.agents.create/list/get/delete/register/unregister/list_connections/get_connection.
• Session and feedback APIs: get_session, add_feedback, delete_feedback, frequency weights.
• Session distillation, feedback-influenced retrieval, and multi-user/agent isolation.

Route away

• Basic permanent-memory storage and recall: core-memory.
• Retrieval mode tuning: search-retrieval.
• Backend/cache/env configuration: configuration-backends.
• CLI/MCP/UI service deployment: api-cli-services.

Read first

1. references/agent-session-memory.md
2. references/api-reference.md
3. references/troubleshooting.md

Safe helper

• scripts/check_agent_memory_payloads.py — validates typed memory payload JSON without writing to a database.

Working rules

• Use a dedicated session_id per agent or task to avoid mixing unrelated traces.
• For feedback updates, keep the qa_id returned by a prior QA memory entry.
• Decorate only async functions with @cognee.agent_memory.
• Resolve dataset access before granting agent permissions.
• Route cache/provider failures to configuration instead of treating them as agent API problems.