为AI提供跨会话、跨模型的持久记忆，解决AI遗忘问题。
AI对话不记忆历史，用户重复信息，工作流断裂，尤其对重度用户和开发者造成效率损失。
重度AI用户、开发者、内容创作者，聚集在Reddit、Product Hunt、AI相关Discord。
AI工具爆发，但记忆缺失成为瓶颈，且MCP协议兴起，为统一记忆层提供技术基础。
跨AI共享记忆，语义去重，矛盾追踪，夜间整合（类似REM），通过MCP无缝集成。
订阅制，个人版$9/月，专业版$29/月，团队版$99/月，提供免费额度。
Python/Node.js后端，向量数据库（如Pinecone），MCP SDK，Stripe。AI可代劳：代码生成、测试、文档。
My name is Filippo Pilotta. For about two years I have been building Cortex. As we all know, AIs are not great at remembering what we wrote in the previous session, or a month ago. I did not build the usual RAG: I built a semantic memory able to govern your AIs' memories, the ones from your conversations, from those long sessions each of us has and wishes every AI would remember. And I pushed it further. Cortex does not just give a memory to a single AI: you can connect it over MCP to all your AIs at the same time, any AI that supports a connector. They will use it natively to save context, there are also skills to improve how you use Cortex, and every AI will have access to the memories of all sessions, including those saved by different AIs. Your work stays available no matter which AI you use. Cortex rejects 80% of writes in production, facts become typed claims, contradictions are tracked instead of overwritten. In practice, if something is already there it does not write it again. And there is a nightly cycle, the REM: yes, like our brain that dreams to reorganize its ideas, Cortex "dreams" too and consolidates its memories. Yesterday it passed OpenAI's review. You can find it