不同工具之間的 AI 記憶問題
The Problem with AI Memory Across Different Tools
ChatGPT、Claude Code 和本地 AI 模型各有用途，但一個工具累積的上下文很少能延續至其他工具，切換時往往要重複説明、重建先前決策，並失去重要脈絡。作者正探索以 Obsidian、Markdown 檔案及 MCP 整合建立供不同 AI 助手存取的共享知識來源，但在檢索相關資訊、更新內容及可靠記錄新知識方面仍面對挑戰。
I use ChatGPT, Claude Code, and local AI models regularly. Each tool serves a different purpose, but they all share one limitation: the context accumulated in one tool rarely follows me to another.
This means repeating explanations, reconstructing previous decisions, and losing valuable context when switching between tools.
I've started exploring how Obsidian, Markdown files, and MCP integrations can help create a shared knowledge source accessible to different AI assistants.
The approach is promising, but challenges remain around retrieving relevant information, keeping it updated, and capturing new knowledge reliably.
In my latest article, I share my experience, the limitations I've encountered, and the approach I'm experimenting with.
Read the full article: When AI Memory Doesn't Follow You Between Tools
How are you managing persistent memory across multiple AI tools?
來源：DEV Community · dev.to