[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"readme:memmolt":3},"\u003Cp>Please read this \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Frituraj-io\u002Fmemmolt\u002Fblob\u002FHEAD\u002Fdocumentations\u002FVERSION1.0.0.md#summary-maintenance\" rel=\"nofollow ugc noopener\">Version file\u003C\u002Fa>\u003C\u002Fp>\n\u003Cpre>\u003Ccode>  ███╗   ███╗███████╗███╗   ███╗███╗   ███╗ ██████╗ ██╗  ████████╗\n  ████╗ ████║██╔════╝████╗ ████║████╗ ████║██╔═══██╗██║  ╚══██╔══╝\n  ██╔████╔██║█████╗  ██╔████╔██║██╔████╔██║██║   ██║██║     ██║\n  ██║╚██╔╝██║██╔══╝  ██║╚██╔╝██║██║╚██╔╝██║██║   ██║██║     ██║\n  ██║ ╚═╝ ██║███████╗██║ ╚═╝ ██║██║ ╚═╝ ██║╚██████╔╝███████╗██║\n  ╚═╝     ╚═╝╚══════╝╚═╝     ╚═╝╚═╝     ╚═╝ ╚═════╝ ╚══════╝╚═╝\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Ch1>MemMolt\u003C\u002Fh1>\n\u003Cblockquote>\n\u003Cp>\u003Cstrong>Structured, searchable, long-term memory for your AI agent.\u003C\u002Fstrong>\nPlugs into Claude Code (and any MCP-compatible client) over the Model Context Protocol.\u003C\u002Fp>\n\u003C\u002Fblockquote>\n\u003Chr \u002F>\n\u003Ch2>What is MemMolt?\u003C\u002Fh2>\n\u003Cp>MemMolt is a memory system for AI agents. Think of it as a brain extension for Claude Code\nthat remembers things across conversations, keeps them tidy, and can recall them by meaning — not just keywords.\u003C\u002Fp>\n\u003Cp>It lives on your machine as a tiny local server. Your agent connects to it, reads from it, writes to it, and picks up where it left off the next time you open a conversation.\u003C\u002Fp>\n\u003Cp>Memory is organized into three simple layers:\u003C\u002Fp>\n\u003Cpre>\u003Ccode>  BUCKET      →    top-level category        (\"Personal Finances\")\n    │\n    └── THREAD   →    sub-topic              (\"Quarterly Taxes\")\n          │\n          └── MEMO   →    the actual note    (\"Q3 estimated payment notes\")\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>That's it. Buckets hold threads. Threads hold memos. Memos are markdown documents.\nNo nested folders, no wiki rabbit holes, no broken links.\u003C\u002Fp>\n\u003Chr \u002F>\n\u003Ch2>What problem does it solve?\u003C\u002Fh2>\n\u003Cp>If you've used an AI agent for a while, you've probably hit one of these:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>\u003Cstrong>\"It forgot what we talked about yesterday.\"\u003C\u002Fstrong>\u003C\u002Fli>\n\u003Cli>\u003Cstrong>\"I keep re-explaining the same context every session.\"\u003C\u002Fstrong>\u003C\u002Fli>\n\u003Cli>\u003Cstrong>\"My notes folder has grown into a jungle I can't navigate.\"\u003C\u002Fstrong>\u003C\u002Fli>\n\u003Cli>\u003Cstrong>\"The agent wastes half its context window reading old notes.\"\u003C\u002Fstrong>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>MemMolt fixes all four. The agent only pulls in the memos it actually needs,\norganized so it knows where to look, indexed so it can find them by \u003Cem>meaning\u003C\u002Fem> and not just word match.\u003C\u002Fp>\n\u003Chr \u002F>\n\u003Ch2>Why not just use Claude Code + Obsidian (or any notes folder)?\u003C\u002Fh2>\n\u003Cp>A plain notes folder works — until it doesn't. Here's what you trade away:\u003C\u002Fp>\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Problem\u003C\u002Fth>\n\u003Cth>Plain notes folder\u003C\u002Fth>\n\u003Cth>MemMolt\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\u003Ctr>\n\u003Ctd>\u003Cstrong>Finding things\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>Text search only. Miss the exact word, miss the note.\u003C\u002Ftd>\n\u003Ctd>\u003Cstrong>Hybrid search\u003C\u002Fstrong>: keyword (FTS5) + meaning (vector embeddings) combined via \u003Cstrong>RRF\u003C\u002Fstrong> so you get both.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>Organization\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>Folders can be nested arbitrarily, links break, stuff drifts.\u003C\u002Ftd>\n\u003Ctd>\u003Cstrong>Enforced 3-level hierarchy\u003C\u002Fstrong>. The agent can't make a mess because the structure doesn't allow one.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>Context window\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>Agent reads multiple full files just to check if they're relevant.\u003C\u002Ftd>\n\u003Ctd>Agent searches summaries first, fetches only the memos it actually needs. Massively fewer tokens.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>Spiraling out\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>Long notes get longer, topics sprawl across files.\u003C\u002Ftd>\n\u003Ctd>Summaries force compression. Each memo has a forced title + summary that must describe what it contains.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>Consistency\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>The agent might remember to update notes, or it might not.\u003C\u002Ftd>\n\u003Ctd>The tool \u003Cem>prompts\u003C\u002Fem> the agent to update summaries after changes. Built-in nudges keep memory fresh.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>Search across topics\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>Manual. You grep, you browse, you hope.\u003C\u002Ftd>\n\u003Ctd>One call returns results from the whole system, ranked by relevance.\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\u003C\u002Ftable>\n\u003Cp>\u003Cstrong>Short version:\u003C\u002Fstrong> A notes folder is a \u003Cem>filesystem\u003C\u002Fem>. MemMolt is a \u003Cem>memory system\u003C\u002Fem>.\u003C\u002Fp>\n\u003Chr \u002F>\n\u003Ch2>How it works (the quick version)\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>Everything lives in \u003Cstrong>one SQLite file\u003C\u002Fstrong> on your disk (default: \u003Ccode>~\u002F.memmolt\u002Fmemmolt.sqlite\u003C\u002Fcode> — see \u003Ca href=\"#configuration\" rel=\"nofollow ugc noopener\">Configuration\u003C\u002Fa> for the full resolution order).\u003C\u002Fli>\n\u003Cli>Summaries are turned into vectors by a local embedding model (\u003Cstrong>all-MiniLM-L6-v2\u003C\u002Fstrong>, runs in-process, no cloud calls).\u003C\u002Fli>\n\u003Cli>Search combines keyword matching (\u003Cstrong>FTS5\u003C\u002Fstrong>) and semantic matching (\u003Cstrong>sqlite-vec\u003C\u002Fstrong>), then merges them with \u003Cstrong>Reciprocal Rank Fusion (RRF)\u003C\u002Fstrong>.\u003C\u002Fli>\n\u003Cli>The agent talks to MemMolt over \u003Cstrong>MCP\u003C\u002Fstrong> (Model Context Protocol) — the same way Claude Code talks to any other tool.\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>No separate database server. N\u003C\u002Fp>\n",1784308502720]