VEKTOR implements concepts from three peer-reviewed papers and one open research project. The implementation is entirely original. No code was copied. These papers shaped how we think about memory.
All concepts explained here are derived from publicly available research. VEKTOR's source code is original and proprietary. Links go directly to the original papers.
Tested on 105 long-context memory recall questions with gpt-4o-mini judge. ONNX MiniLM-L6-v2 embedder. RRF dual-channel retrieval (vector + BM25).
LongMemEval v4 benchmark. 105 long-context memory questions across 7 categories.
gpt-4o-mini evaluates whether recalled context answers each question correctly.
ONNX all-MiniLM-L6-v2. 384-dimensional semantic vectors. Local execution.
SQLite + pgvector. Identical accuracy on both. Same RRF retrieval algorithm.
RRF dual-channel fusion. Vector (pgvector HNSW) + BM25 keyword search ranked together.
June 22, 2026. Validated with extended test harness. Production-ready.
Four systems working together: the ingest pipeline that decides what to write, multi-signal retrieval that fuses three search channels, the MAGMA 4-layer graph, and the AUDN conflict resolution loop.
Every memory.store() call probes for related context, runs AUDN to decide the verdict, embeds and links entities, then writes to the graph. NO_OP skips the write entirely — graph stays lean.
Three channels run in parallel and fuse results via Reciprocal Rank Fusion. BM25 is weighted 2× — keyword precision outperforms semantic alone for technical queries. Combined score beats any single channel.
Four distinct memory layers, each answering a different question. Recall traverses all four simultaneously — entity, semantic, causal, and temporal signals combine into a ranked result that pure vector search cannot produce.
The AUDN loop fires before every write. Context probe finds the top-5 related nodes via BM25 + semantic search, then routes to the correct verdict. Four possible outcomes keep the graph accurate without any manual curation.
VEKTOR Slipstream is the memory core. Four tools in the ecosystem each solve a distinct layer of the agent infrastructure problem.
Cross-standard vector database migration and memory portability. Move your entire memory graph between any vector store in one command. Import Claude and ChatGPT conversation history with LLM fact extraction. Back up to any Git host with AES-256-GCM client-side encryption.
npm install -g @vektormemory/vex
vex sync pullMCP configuration sync. One .mcp.json file. Every AI editor. Always in sync. Define your MCP servers once and push to Claude Desktop, Cursor, VS Code, Windsurf, Claude Code, Cline, Roo Code, Gemini CLI, GitHub Copilot, Continue, and Codex with one command.
npm install -g @vektormemory/vek-sync
vek-sync init --from cursor — pulls your current config as the source of truth--dry-run — see exactly what changes in each editor configvek-sync ping — spawns each server and checks MCP handshakevault:key-name — never written to editor configs as plaintextvek-sync share — secrets stripped before upload, teammates pull with one command.mcp.json and installed editors with vek-sync diff — exits non-zero in CIUniversal AI tool integration layer. Self-improving prompt engine that learns from your codebase history. Every correction stored. Every success a reusable pattern. Works with Claude Code, Codex, Gemini CLI, Cursor, and Windsurf.
npm install -g @vektormemory/via
Cryptographic proof of what your code looked like, and when. Stamps proprietary headers into your source, snapshots the codebase into a Merkle-tree manifest bound to your git commit, and anchors that manifest with two independent timestamp authorities so no single authority is a point of failure.
npm install -g @vektormemory/prov
--check modeLocal-first note-taking app with persistent AI memory built in. Every note stored in the VEKTOR memory graph automatically. JOT Collab fires four seconds after you stop typing — surfaces an insight, a gap suggestion, and four arXiv papers from the literature, without leaving the editor.
Google Play — Download FreeEvery piece of information is stored as a node in a typed, attributed graph. Semantic nodes capture facts. Causal nodes capture why. Temporal nodes track when. Entity nodes track who. Edges between them encode relationships, not just proximity.
Before writing any memory, VEKTOR asks the model: does this already exist? Does it contradict something? Should it replace a previous belief? The AUDN decision (Add / Update / Delete / None) fires on every remember() call. The graph never accumulates stale data.
Inspired by EverMemOS lifecycle management. While the agent idles, a 7-phase background process scans recent memories, identifies clusters, and compresses fragments into higher-level insights. The result: the graph grows wiser, not just larger. Available in VEKTOR Slipstream.
Unlike keyword RAG, memory.recall() traverses the graph using vector similarity and edge weights together. A query doesn't just find matching text — it finds connected context: the reason something was remembered, the time it changed, and the entities involved.
Pure SQLite. No cloud sync, no external API calls for memory operations. The graph lives in a single .db file on your server. Backup is a file copy. Migration is a file move. Memory never touches a third-party service — your agent data is yours.
The MemGPT OS paradigm made practical. Each session begins with memory.briefing() — a structured summary of what the agent knew when it last ran. Topics reviewed, decisions made, open threads. The agent doesn't start from zero; it wakes up and continues.
Every line in VEKTOR was written from scratch. The papers above shaped how we think about memory architecture — they are the intellectual foundation. But the implementation decisions: the SQLite schema, the AUDN prompt design, the REM phase sequence, the embedding pipeline — those are original VEKTOR engineering.
Publishing the research foundations lets you verify the concepts behind VEKTOR. If you're evaluating whether VEKTOR's approach is sound, read the papers. If you want to understand the implementation, full documentation is included with your purchase.
From $9/month. Local-first. Drop into any Node.js agent in minutes.