Give Your AI Agents Persistent Memory with Memsearch

𝗟𝗟𝗠𝘀 𝗮𝗿𝗲 𝗲𝘃𝗼𝗹𝘃𝗶𝗻𝗴 𝗳𝗮𝘀𝘁. 𝗬𝗼𝘂𝗿 𝗮𝗴𝗲𝗻𝘁𝘀 𝗺𝗮𝘆 𝗰𝗵𝗮𝗻𝗴𝗲 𝗲𝘃𝗲𝗿𝘆 𝘄𝗲𝗲𝗸, 𝗮𝗻𝗱 𝘁𝗵𝗲 𝗺𝗲𝗺𝗼𝗿𝘆 𝗳𝗿𝗼𝗺 𝗽𝗮𝘀𝘁 𝗶𝗻𝘁𝗲𝗿𝗮𝗰𝘁𝗶𝗼𝗻𝘀 𝗱𝗶𝘀𝗮𝗽𝗽𝗲𝗮𝗿𝘀 𝘄𝗶𝘁𝗵 𝘁𝗵𝗲𝗺. Every migration means reloading context, repeating details, and hoping nothing important gets lost. More importantly, we want memory that we 𝗰𝗮𝗻 𝗶𝗻𝘀𝗽𝗲𝗰𝘁 𝗮𝗻𝗱 𝗲𝗱𝗶𝘁 𝗮𝘁 𝗮𝗻𝘆 𝘁𝗶𝗺𝗲, 𝗻𝗼𝘁 𝗮 𝗯𝗹𝗮𝗰𝗸 𝗯𝗼𝘅. 𝗧𝗵𝗮𝘁 𝗶𝘀 𝗲𝘅𝗮𝗰𝘁𝗹𝘆 𝘄𝗵𝘆 𝘄𝗲 𝗯𝘂𝗶𝗹𝘁 𝗺𝗲𝗺𝘀𝗲𝗮𝗿𝗰𝗵, an open-source AI memory system built on Milvus, with plug-and-play Claude Code plugin. On top of that, memsearch also offers: • 𝗢𝗽𝗲𝗻𝗖𝗹𝗮𝘄'𝘀 𝗺𝗲𝗺𝗼𝗿𝘆, 𝗲𝘃𝗲𝗿𝘆𝘄𝗵𝗲𝗿𝗲 -- markdown as the single source of truth • 𝗛𝘆𝗯𝗿𝗶𝗱 𝘀𝗲𝗮𝗿𝗰𝗵 — dense vectors + BM25 + RRF reranking • 𝗦𝗺𝗮𝗿𝘁 𝗱𝗲𝗱𝘂𝗽 — SHA-256 hashing, so unchanged content is never re-embedded • 𝗟𝗶𝘃𝗲 𝘀𝘆𝗻𝗰 — automatically indexes changes and removes stale chunks • 𝗠𝗲𝗺𝗼𝗿𝘆 𝗰𝗼𝗺𝗽𝗮𝗰𝘁 — LLM-powered summarization compresses old memories