Your sovereign AI workspace
Research, your library, recordings, notes, chat and workflows, interconnected inside a firewall of light, while the open internet stays on the outside until it is safe.
Autonomous research, every claim traceable to its source.
Your documents, AI-usable yet unreadable to anyone but you.
An EU-only data path, with keys held on your device.
We cannot read your data. Not by policy, by architecture.
It fills itself. Submit a question and LumaVista does the gathering — a planner decomposes it, specialist agents search, reason, and synthesize across a directed graph, and every claim in the final report is traced back to its evidence. The cold-start problem every knowledge tool has, solved.
Bring your private knowledge — and talk to all of it. Internal documents, notes, and recordings join the same library, made AI-usable while staying encrypted and unreadable to anyone but you. Chat grounds across your whole library and the web at once, with scope selection so you decide exactly what each answer draws on.
It runs your playbook — privately by architecture. Teach it a procedure once and it runs on every new matter, behind a boundary that treats the outside world as hostile. EU-sovereign infrastructure, device-held keys, and a server that structurally cannot read your data. Works with any OpenAI-compatible LLM, or pair it with Meridian for private inference on your own GPU fleet.
One platform. Three ways to run it.
The same engine, the same security model — hosted by us, isolated for you, or entirely on your own infrastructure. You choose where it lives; nothing about the product changes.
Managed SaaS
EU-hosted at app.lumavista.eu. The fastest start — per-seat, fully managed; we run the ops and the updates. Your data stays in the EU and your keys stay device-held.
Dedicated cloud
A single-tenant instance that is yours alone — dedicated database, storage, and backend, in the region you choose. Hard isolation without running the hardware yourself.
On-premise
Runs entirely inside your boundary. Docker Compose, air-gap capable. Research data never leaves your network, and GDPR deletion is a single rm -rf data/<userID>/.
What it does
Autonomous research
A multi-agent DAG engine. A planner decomposes your question; specialist agents — search, reasoning, aggregation, writing — execute in parallel across the graph. Configurable depth, budget guards, and human-in-the-loop checkpoints. Every source and decision is inspectable.
Your private library
Upload documents, notes, and recordings — PDF, DOCX, HTML, Markdown — into one library. Converted and indexed with local embeddings (no external API), encrypted per-user, and unreadable to anyone but you.
Talk to all of it
Chat grounded across your whole library and the live web at once — with scope selection, so you decide exactly what each answer is allowed to draw on. Cited, traceable, yours.
Runs your playbook
Event-triggered workflows and connector-skills run your standard procedures on every new matter — behind a boundary that filters every inbound byte before it touches your knowledge and guards every outbound action. Your expertise, on autopilot.
Living memory
A persistent knowledge graph, auto-extracted from your results. Entities, topics, and relations accumulate and are reused across projects — the platform gets sharper with every investigation, and so do you.
Private search
A self-hosted SearXNG meta-search engine — no query ever leaves your infrastructure. Local page retrieval and caching, with source-credibility scoring on every result. No third-party search API.
Built for regulated teams
Identity, governance, and compliance are first-class — not add-ons. Independence at every layer: your infrastructure, your keys, your judgment. Never ours.
SSO & SAML
Organization-level identity with domain-based routing and an enforced sso_required login gate. Bring your own IdP.
Organizations & RBAC
Organizations, memberships, and roles spanning system and org scope, with per-seat management. The right people see the right surfaces.
Audit logs & admin
A full audit trail, an admin dashboard, and runtime feature controls — flip capabilities live without a redeploy.
Right to erasure
Self-serve and operator-driven account deletion that crypto-shreds keys and cascades across every store in order. GDPR/DSAR by construction.
Data residency & sovereignty
An EU-only data path. The sovereign trust tier degrades to private inference, fail-closed — never silently routing your data to an external model.
Encryption & key control
Device-controlled keys and per-user isolation. The server structurally cannot read your data — not a policy promise, an architectural one.
Billing & cost governance
A per-seat and credit model with organization budgets and budget guards. Cost is communicated in credits — never a surprise invoice.
Connectors
OAuth2 connectors for Slack, Google, and GitHub plus MCP — each behind the same inbound-filter and outbound-guard boundary as everything else.
Technical specifications
| Backend | Go |
| Frontend | Next.js 15, React 19, TypeScript |
| Protocol | Binary protobuf over WebSocket (real-time), REST (management) |
| Storage | Per-user Badger DB (user data), PostgreSQL (system), Redis (ephemeral) |
| Authentication | WebAuthn / passkeys (no passwords), SSO / SAML for organizations |
| Encryption | Device-controlled keys, per-user isolation, GDPR deletion = rm -rf data/<userID>/ |
| LLM support | Any OpenAI-compatible API, native Meridian integration |
| Deployment | Managed SaaS (EU), dedicated single-tenant cloud, or on-premise (Docker Compose, multi-backend) |
| Search | Self-hosted SearXNG, crawl4ai, markitdown |
| Vector search | Local embeddings — no external API |
Works with Meridian
LumaVista works with any OpenAI-compatible LLM endpoint out of the box. Pair it with Meridian and your research agents gain capability-based routing, priority queuing, and GPU fleet auto-scaling — all on your own infrastructure. The integration is transparent: point LumaVista's LLM base URL at your Meridian gateway and the platform handles the rest.
See it in action
A guided walkthrough of LumaVista — screenshots, diagrams, and the thinking behind each feature. Best way to understand the product before you talk to us.
Take the tour
