Unverified50% confidenceFactExact time
By default, the agent starts from scratch in each new session and does not retain history, which leads to the need to repeatedly provide the same feedback; memory files solve this problem by writing information that is valid across sessions to disk and reading the context back at the next startup.
1
Sources
50%
Confidence
Long-term
Relevance
11/1/2026
First Seen
Sources
Build Your First AI Agent with Claude Code in 20 Minutes — No Code Required
youtubeGrace Leung10/30/2026
Related Entities
Related Claims
UnverifiedBackboard provides a persistent memory layer across sessions, writing the facts generated by each run into long-term storage and recalling them in subsequent runs, which is different from one-off retrieval RAG.75% similarUnverifiedTwo default practices for long-term memory: extraction should be placed asynchronously outside the reply path (because the facts of this conversation are already in this session's history); take a snapshot of memory at the start of each conversation (the memory block is placed before the history, and updating it midway would break the cache).73% similarVerifiedWhen new information conflicts with old memories, the memory layer must be able to identify and update them rather than holding on to outdated knowledge.73% similarUnverifiedMemories saved via /memory are automatically retained in every future session without requiring manual maintenance.73% similarUnverified第一天做记忆时，一个markdown文件（如preferences.md）写下偏好规则就胜过大多数花哨的记忆系统，因为可直接读、直接改；每次会话把它加载进系统提示词；等文件超过约2000词再考虑加检索。73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1122569API
curl https://kongchang.com/api/v1/knowledge/claims/1122569MCP
get_claim(id=1122569)