Documentation · Vektor Slipstream v1.8.0

Build with
persistent memory.

Everything you need from first install to production deployment. API reference, integration guides, and troubleshooting in one place.

Get Started
Purchase a licence key to download the CLI
$9/month · Cancel any time
Buy Now
Version
v1.8.0 · latest
Node
≥ 18.0.0
Size
23MB · ONNX bundled
docs/assistant.md
WORKERS AI 10 / hr
1
2
3
4
5
6
7
8
9
quick help
>_
0/10
wizard.getvektor.workers.dev
UTF-8 · llama-3.3-70b
Documentation · 12 sections
01 · Start Here
Quickstart

Purchase to first memory stored in under 5 minutes. The fastest path from zero to a working agent with persistent memory.

Install the package
Activate your licence key
Store and recall your first memory
Understand the boot sequence
Get started →
02 · Reference
API Reference

Complete documentation for every method parameters, return types, and working examples for each one.

createMemory()
remember() · recall() · graph()
delta() · briefing() · dream()
Performance notes and best practices
View API →
03 · Integrations
Integration Guides

Step-by-step guides for every supported framework. Copy, adapt, ship. Full example files included in the package.

LangChain persistent researcher agent
OpenAI Agents SDK memory tool loop
Claude MCP Desktop + direct chat
Mistral HMAC-bound vektor_memoire
Pick your stack →
04 · Support
Troubleshooting

Common errors with exact fixes. Licence issues, ONNX problems, SQLite conflicts diagnosed and resolved.

Licence validation errors
ONNX model not found
SQLite BUSY / locked database
Module not found errors
Fix issues →
CLI · vektor
05
CLI Reference
Every vektor command setup, TUI, MCP, REM, briefing. Full flag and environment variable reference.
  • vektor setup · activate · status · test
  • vektor mcp start Claude MCP server
  • vektor rem · briefing
View CLI docs →
DXT · Claude Desktop
06
Claude Desktop Extension
Install the .dxt extension to connect Claude Desktop to persistent memory. One drag-and-drop no config files, no JSON editing.
  • Download vektor-slipstream.dxt
  • Drag onto Extensions page in Claude Desktop
  • Enter licence key 34 MCP tools available instantly
  • vektor_recall · vektor_store · vektor_graph · vektor_delta
View DXT docs →
CLOAK · Stealth browser · Vault
07
Cloak Stealth Browser & Vault
Bot-resistant headless browser, AES-256 credential vault, layout sensor, and self-improving behaviour patterns. Available as MCP tools and SDK methods.
  • cloak_fetch · cloak_diff · cloak_render
  • cloak_passport AES-256-GCM encrypted vault
  • cloak_identity_* persistent browser fingerprints
  • cloak_warmup_session · cloak_inject_behaviour
  • turbo_quant_compress 10x vector DB compression
View Cloak docs →
NEW · v1.7.7
08
Memory Graph Dashboard

Visualise your agent’s memory as a live D3 force graph. Filter by semantic, causal, temporal, or entity nodes. Three themes. Opens with /graph.

/graph · /dash · /dashboard
Graph · Health · Settings tabs
147+ nodes · real-time sidebar count
Auto-starts on boot · port 3847
View Dashboard docs →
NEW · v1.7.7
09
Settings GUI Config Manager

Set providers, API keys, and models from the browser Settings tab. No JSON editing. Hot-reloads into CLI in real time.

Visual model picker all 16 providers
API Keys: SET / UNSET / ENV badges
Export .env file with one click
Config hot-reload 2 second sync
View Settings docs →
NEW · v1.7.7
10
Memory Intelligence

Confidence scoring, dedup, recall tuning, RL prioritisation, briefing scheduler, self-organising Zettelkasten memory, spec-decoding retrieval, and query-anticipation ingestion.

vektor-confidence · vektor-dedup
vektor-recall-tune · vektor-rl-memory
vektor-briefing-scheduler · vektor-selforg
vektor-session-ingest · slipstream-embedder v1.2
Transparent proxy chain · fails open
View Intelligence docs →
NEW · v1.7.7
11
Import & Merge

Drop-folder DB consolidation, schema migrations, namespace isolation, memory export, and MAGMA causal diff.

Drop .db into ./import auto-merges
Export: json · markdown · csv
Namespace isolation · cross-namespace recall
MAGMA causal lineage reconstruction
View Import docs →
NEW · v1.7.7
12 · Infrastructure
SSH Tools

Execute commands, deploy files, and manage remote servers directly from Claude Desktop or any MCP agent. Built-in safety tiers, auto-backup, and one-step rollback.

cloak_ssh_exec run any shell command
cloak_ssh_plan multi-step atomic deployments
cloak_ssh_approve safety gate with auto-backup
cloak_ssh_rollback one-step undo
View SSH docs →
NEW · v1.7.7
13 · GUI
JOT Notes & Synthesis

Split-panel note interface with live streaming AI synthesis, clickable insight expansion, named session management, arxiv research panel, and built-in unit converter & math evaluator.

Streaming synthesis 600ms debounce, live token render
Clickable circles expand any insight inline
Named sessions save, list, reopen, delete
AI toolbar summarise, flashcards, whitepaper, actions
Unit converter length, weight, temp, currency
Arxiv panel live related research below synthesis
v1.0 · v1.7.8
14 · Reference
Changelog

Full version history from v1.0.0 to v1.8.0 every new tool, module, and fix across seven major releases.

v1.7.7 MAGMA graph, status & related tools
v1.4.0 intelligence modules, namespace isolation
v1.3.7 TypeScript types, audit log, 13 security fixes
v1.0 → v1.2 hybrid recall, namespaces, initial release
View changelog →
USE CASE · v1.7.7
15
Claude Code & Cursor

Persistent memory for AI coding agents survives session resets and context compaction. MCP-native setup in one command.

Architectural decisions remembered across sessions
Survives context compaction
Shared graph across both editors on one machine
View guide →
USE CASE · v1.7.7
16
Multi-Agent Systems

One shared memory graph across every agent in your pipeline. CrewAI, LangChain, and n8n adapters with AUDN conflict resolution.

Five-verdict conflict resolution, not last-write-wins
Trust-weighted human vs. agent writes
CrewAI · LangChain · n8n adapters
View guide →
USE CASE · v1.7.7
17
Personal Knowledge Work

A private, local-first memory graph for your own notes, research, and writing. JOT capture, REM compression, auto-wired connections.

Auto-wired SUPPORTS / EXTENDS / CONTRASTS edges
REM cycle compresses fragments into core insights
VEKTOR Notes companion app for Android
View guide →
SECURITY · v1.7.5 Preview
18
Faraday-Gate MCP Security Proxy

A transparent proxy that scans every MCP tool schema and call before anything reaches VEKTOR memory. Tool pinning, canary tokens, taint propagation, human approval gate.

SHA-256 tool pinning catches supply-chain rug-pulls
Gate queue holds high-risk actions for review
faraday_status · faraday_approve_action
View Faraday-Gate docs →
FEATURE · v1.7.7
19
Memory Transparency

Every memory visible, searchable, and editable in the same interface you already work in. Consolidation proposes a rewrite nothing saves until you approve it.

Sequenced search a slow result can never overwrite a newer one
Consolidation is fully supervised, never auto-saved
Atomic batch delete exact selection, no partial state
View Memory Transparency docs →
FEATURE · v1.8.0
20
Agent Tab Full IDE Layout

A 3-column IDE layout for VEKTOR's tool-using agent: chat thread, file-tree / diff-viewer / docked terminal, and a live memory-recall rail, all in one panel.

RUN / SEND composer live chat interrupts an active run instead of queuing a second one
Per-tool allow/ask/deny gating with an in-panel approval modal
Write self-check every write is syntax-validated the moment it happens
View Agent Tab docs →
FEATURE · v1.8.0
21
CodeGraph 36-Language Code Intelligence

On-demand, session-scoped code understanding no daemon, no persistent index. Real AST-level parsing now spans 36 languages, each fetched and hash-verified only the first time it's actually needed.

36 languages, fetched on first use not bundled upfront
Every grammar SHA-256 verified against a pinned manifest before it's trusted
10 core languages pre-warm on install; everything else stays out of the way until needed
View CodeGraph docs →