๐ช๐ผ๐ฟ๐ฟ๐ถ๐ฒ๐ฑ ๐๐ผ๐'๐น๐น ๐น๐ผ๐๐ฒ ๐๐ผ๐๐ฟ ๐๐น๐ฎ๐๐ฑ๐ฒ ๐๐ผ๐ฑ๐ฒ ๐บ๐ฒ๐บ๐ผ๐ฟ๐ ๐ถ๐ณ ...

- memsearchๅฐAI่ฎฐๅฟๅญๅจไธบๅฏ็ผ่พ็Markdownๆไปถ
- ้่ฟMilvus Liteๅฎ็ฐๆฌๅฐ็ดขๅผ๏ผๆฏๆ่ฏญไนๆฃ็ดข
- ๆฐๆฎๆๆๆๅฝ็จๆท๏ผๅฏ้ๆถ้ๅปบ็ดขๅผๆๅๆขๅนณๅฐ
The ๐บ๐ฒ๐บ๐๐ฒ๐ฎ๐ฟ๐ฐ๐ต ๐ฝ๐น๐๐ด๐ถ๐ป for Claude Code keeps all your important memory on your own disk.
Last week, https://t.co/GHeRUNuF32" / X
Post
Conversation
๐ช๐ผ๐ฟ๐ฟ๐ถ๐ฒ๐ฑ ๐๐ผ๐'๐น๐น ๐น๐ผ๐๐ฒ ๐๐ผ๐๐ฟ ๐๐น๐ฎ๐๐ฑ๐ฒ ๐๐ผ๐ฑ๐ฒ ๐บ๐ฒ๐บ๐ผ๐ฟ๐ ๐ถ๐ณ ๐๐ป๐๐ต๐ฟ๐ผ๐ฝ๐ถ๐ฐ ๐๐๐๐ฝ๐ฒ๐ป๐ฑ๐ ๐๐ผ๐๐ฟ ๐ฎ๐ฐ๐ฐ๐ผ๐๐ป๐? ๐๐ฒ๐ฎ๐ฟ ๐ป๐ผ๐. The ๐บ๐ฒ๐บ๐๐ฒ๐ฎ๐ฟ๐ฐ๐ต ๐ฝ๐น๐๐ด๐ถ๐ป for Claude Code keeps all your important memory on your own disk. Last week, Anthropic banned more than 60 of Belo's accounts overnight. Belo's CTO Pato Molina posted the receipt: x.com/patomolina/sta Belo had Gemini pre-deployed as backup and kept running, but everything they'd built inside Claude was stuck on Anthropic's side. ๐บ๐ฒ๐บ๐๐ฒ๐ฎ๐ฟ๐ฐ๐ต ๐๐ผ๐๐น๐ฑ ๐ต๐ฎ๐๐ฒ ๐ฝ๐ฟ๐ฒ๐๐ฒ๐ป๐๐ฒ๐ฑ ๐ฎ ๐น๐ผ๐๐ ๐น๐ถ๐ธ๐ฒ ๐๐ต๐ฎ๐. ๐บ๐ฒ๐บ๐๐ฒ๐ฎ๐ฟ๐ฐ๐ต can save Claude Code memories as plain markdown files on your own disk, not in Anthropic's database. Index it locally with Milvus Lite, or scale to Zilliz Cloud when your team grows. Whatever happens, the memory will belong to you. ๐๐ผ๐ ๐ถ๐ ๐๐ผ๐ฟ๐ธ๐: most memory tools lock your data in a vector database you can't read or edit. memsearch inverts that. Memories get automatically written to markdown files first, and a vector index sits on top for fast semantic retrieval. The markdown is the source of truth. If the index gets out of sync, rebuild it from the files. If the AI remembers something wrong, edit the file by hand and the index re-syncs. memsearch ships as a Python library too, so any agent you build can plug in, not just Claude Code. ๐ฆ๐๐ฎ๐ฟ๐ ๐๐ถ๐๐ต ๐๐ต๐ฒ ๐๐น๐ฎ๐๐ฑ๐ฒ ๐๐ผ๐ฑ๐ฒ ๐ฝ๐น๐๐ด๐ถ๐ป: milvus.io/blog/adding-pe ๐๐ถ๐๐๐๐ฏ: github.com/zilliztech/mem
