v0.1.7 · Smarter graph · MCP /update /status /compile · Windows + macOS live

Claude forgets you after every conversation. NeuroVault doesn't.

A local-first memory layer for AI agents. Notes, decisions, people, projects — captured once, recalled whenever any MCP-speaking agent asks. Analytics mode shows the structure of your brain at a glance: bigger nodes for what gets referenced most, soft tints for clusters of linked notes.

90%
hit@5 on
long-memory bench
0 ms
cloud round-trip
(fully local)
9.8 MB
installer
(was 76 MB)
1-click
install
& Claude integration
THE PITCH

Most "agent memory" products are a RAG pipeline in a trench coat.
NeuroVault is a living wiki your agent can read, write, and challenge.

Every memory is a plain .md file on your disk. The database is just an index. If the index ever breaks, we rebuild it from the files. You own your brain.

WHAT YOU GET

Memory that sticks.

 ╭─╮
 │ │
╭┴─┴╮
│ ● │
╰───╯

Never leaves your laptop

No cloud. No account. No telemetry. Just files.

●───●
│╲ ╱│
│ ● │
│╱ ╲│
●───●

See how you think

Every note is a node. Force-directed graph, typed links, live.

╭───╮
│ ● │
╰─┬─╯
  ╲
   ╲

Finds anything, fast

Semantic + keyword + graph search, fused. Weighted by recency.

╭─╮
│█│
│█│
│█│
 ▼

Write without friction

Live preview, [[wikilinks]], slash commands, 7 themes.

┌─────┐
│ +   │
│ -   │
│ + ✓ │
└─────┘

AI drafts, you approve

Claude synthesizes wiki pages from raw notes. Every change is a diff.

╭─╮╭─╮╭─╮
│W││R││P│
╰─╯╰─╯╰─╯
 ●  ○  ○

A brain per context

Work, research, personal — one keystroke. Obsidian-compatible.

╭────╮
│┌──┐│
│└──┘│
│████│
╰────╯

Claude remembers everything

One JSON in your agent config. Persistent memory, every chat.

╲ │ ╱
 ╲│╱
──✦──
 ╱│╲
╱ │ ╲

Thinks like you do

Recall walks the graph — finds what you meant, not just what you typed.

  ●
 ╱│╲
● │ ●
 ╲│╱
●─●─●

See your brain's structure

Analytics mode: bigger nodes for what gets referenced most (PageRank), soft tints for clusters of linked notes (Louvain). Tip bar explains each node's role on hover.

┌──┐
│⌘ │
│/ │
└──┘
 ✦

Your AI session names your clusters

No API keys. The /name-clusters skill runs in your existing Claude / Cursor session via MCP. Same pattern coming for deduplication, folder filing, and frontmatter lint.

SEE IT

Two views, one memory.

NeuroVault neural graph — force-directed knowledge graph
Neural graph. Force-directed layout. Default view: nodes coloured by folder, edges thicker for stronger relationships (manual [[wikilinks]] bolder than weak semantic matches). Flip on Analytics mode and nodes resize by importance, soft background tints group communities, and a tip bar explains each node's role.
NeuroVault command palette — commands, notes, and semantic memory in one prompt
Command palette. ⌘K is the primary nav verb. Three sections in one palette — Commands (fuzzy), Notes (title search), Memory (semantic recall after three characters).
HOW IT WORKS

Three paths to the same memory.

01

You write a note

Auto-saved as markdown. A file watcher triggers the ingestion pipeline — chunk, embed locally, extract entities, update the knowledge graph.

02

You drop a fact in chat

A UserPromptSubmit hook silently runs your message through an extractor. Eight patterns catch preferences, decisions, deadlines, stacks. Each fact becomes a first-class kind='insight' engram with a wikilink back to where you said it.

03

The agent asks

Claude calls recall() via MCP. Hybrid search fuses semantic, keyword, and graph hits. Cross-encoder reranks the top candidates; PageRank importance lifts hub notes when Analytics mode is on. Answers now carry context they couldn't have had before.

DATA SAFETY

100% local. Actually.

Give Claude a memory that survives the window close.