[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$ffskn7uvro6b1":3},{"code":4,"message":5,"data":6,"timestamp":58},10000,"success",{"full_name":7,"position":8,"description":9,"language":10,"stars":11,"forks":12,"stars_gained":13,"age_days":14,"hn_threads":15,"one_liner":26,"summary_md":27,"topic":28,"github_topics":31,"license":52,"homepage":53,"updated_at":54,"history":55},"thedotmack\u002Fclaude-mem",5,"Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More","TypeScript",102769,9078,274,418,[16,22],{"id":17,"title":18,"points":19,"comments":20,"posted_at":21},47558167,"Claude-Mem",2,1,"2026-03-28",{"id":23,"title":18,"points":20,"comments":24,"posted_at":25},46229436,0,"2025-12-11","让编程 Agent 跨会话保留记忆，不必每次重开都要重新学习你的代码库。","## 解决什么问题\n\n每开一个新会话，Agent 都是从零开始：你要重新解释架构、重讲昨天已经定位好的 bug、再重申一遍刚定下的约定，Agent 也会重新翻一遍同样的文件去\"发现\"双方早就知道的事。上下文窗口只能在单次会话内帮忙，跨不过会话边界。Claude-Mem 的做法是在会话过程中捕获 Agent 的工具调用，压缩成观察记录，再把相关内容注入之后的会话。\n\n## 有什么不同\n\n宿主工具通常只提供超长上下文或者一个项目级指令文件，靠人手动维护，静态且很快过时。通用 Agent 框架往往要求你自己搭记忆层，而通用记忆服务（Chroma\u002FSQLite 存储、托管记忆 API）并不挂在编程 Agent 的生命周期上。Claude-Mem 以插件加 worker 的形态接进多个 Agent，采用分层渐进式检索，能看到每层检索消耗多少 token，本地用 SQLite 加向量做索引。真正拉开差距的是 hook 级别的接入和 token 成本透明；抽取本身是 LLM 摘要管线，质量就是产品的全部。\n\n## 适合谁\n\n适合在同一仓库上连续数周反复开会的开发者，尤其是用 Claude Code 或兼容 Agent 的人（README 列了 OpenCode、Codex、Gemini、Copilot、Grok Bot 等），也适合想在个人记忆之上再叠一个团队共享工作区的团队。只用一次性对话、处在严格数据外发管控环境、或者已有自己信得过的记忆方案的人，可以先跳过——自动注入的记忆层是多一个可能悄悄出错的地方。\n\n## 上手成本\n\n需要 Node 20 及以上，还要跑一个 worker 服务。默认安装流程会先引导你登录托管账号，给 30 天试用，到期后回落到你自己的 Anthropic 套餐或 OpenRouter\u002FGemini key；免费额度的说法和一长串集成列表值得仔细读。纯本地模式需要显式开启，Grok Bot 这条路是直接读磁盘上的聊天日志。Apache-2.0 许可、TypeScript 实现，迭代非常快（一年半不到就到了 15.x）。升级时预计会碰到破坏性变更，记忆召回的内容也可能和代码现状对不上。\n\n## 点评\n\n方向是对的，而且晚了：Agent 编程的瓶颈在连续性，不在模型本身。但很难说这是成熟项目还是一次尖峰：102,726 star，却只在日榜露面一天（当日 +274），当月新增 +8,445，说明总量是在短时间内冲上去的，而不是稳定增长。这更像一个踩中风口、短期内爆发的项目。代码库显然在活跃维护，但自动注入的记忆会在你不知情的情况下改写每个 prompt，建议先在试验仓库里试，看清它到底召回了什么，再放心用到正经项目上。",{"slug":29,"name":30},"developer-tools","开发工具",[32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51],"ai","ai-agents","ai-memory","anthropic","artificial-intelligence","chromadb","claude","claude-agent-sdk","claude-agents","claude-code","claude-code-plugin","claude-skills","embeddings","long-term-memory","mem0","memory-engine","openmemory","rag","sqlite","supermemory","Apache-2.0","https:\u002F\u002Fclaude-mem.ai","2026-10-24T15:06:04Z",[56],{"date":57,"stars":11,"stars_gained":13},"2026-10-24",1792868616]