It notices what matters
A letter, contract or meeting arrives. Zenith identifies the subject, people, dates and obligations.
Private by design · Agentic by choice · Built on your documents
Your archive.
Your private AI.
Ready to act.
Stop searching through folders and rebuilding context by hand. Zenith turns documents and meetings into a private AI work partner that finds what matters, shows its sources and prepares the next step – on hardware you control.
Working document prototype · Agent and meeting transcription in development
Works fully offline
Cited answers
Your data. Your control.
128GB
Unified RAM for AI
0Bytes
Cloud transfer in the core workflow
3×
Specialised AI models
<30min
Setup time
Why Zenith exists
Your archive should not just store information.
It should work for you.
A scanned letter should not create another evening of renaming files, searching folders and reconstructing context. Zenith is built to understand what arrived, connect it to what you already know and help you move forward.
- 01
Bring order to the past. Turn scans, files and correspondence into a connected archive.
- 02
Understand the present. Find the right facts, people, deadlines and decisions with sources.
- 03
Prepare what comes next. Let your private AI partner propose the next useful action.
2:26 min · See the working Zenith prototype
Everything at a glance
What Zenith does for you
Multimodal AI capture – reads PDFs, Office files, scans and photos, including tables and handwriting
Semantic search – finds documents by meaning, not just keywords
AI chat with sources – ask questions in natural language, get answers with document references
Knowledge graph – interactive view of people, companies, topics and their documents
Automatic calendar – detects appointments and deadlines in documents and reminds you in time
Folders & export – organise documents in virtual folders and export them with one click, e.g. for your tax advisor
Local by default – core AI processing stays on your device; external tools run only when you explicitly enable them
Unlimited users – web app in the local network, no per-user licences
Users & permissions – role-based access is enforced consistently across documents, search, AI chat and knowledge graph
Audit trail – important access, exports, changes and administrative actions remain traceable
Backup & recovery – automatic local backups, visible backup status and a controlled restore process protect your archive
Administration & updates – system health, storage, queues, models, backups and offline updates are managed centrally
Zenith Agent · in development – plans document tasks, uses approved tools and asks before consequential actions
Meeting transcripts · in development – creates local transcripts, summaries, decisions and action items
The problem with cloud AI
Your data is your capital.
Why hand it over?
Your archive holds contracts, conversations and decisions that should not become somebody else's dataset. Zenith keeps its core document and AI workflow local; optional external tools are used only when you explicitly enable them.
⚠ Cloud AI
The invisible risk
- ✕Sensitive business data and trade secrets end up on third-party servers – often outside the EU
- ✕Potential violations of GDPR, trade-secret law and industry-specific compliance
- ✕Opaque storage: training data, logs and backups beyond your control
- ✕Monthly per-user fees that never stop – costs grow with your team
- ✕No internet, no AI: your workflows depend on someone else's uptime
✓ Zenith
Local sovereignty
- ✓Core AI processing runs on your hardware; external tools remain optional and explicit
- ✓Privacy-by-design: local processing supports GDPR-compliant operation and audits
- ✓Full data control: you decide what is stored, processed and deleted
- ✓One-time purchase – no subscription, no per-user licences, unlimited usage
- ✓Works fully offline – even air-gapped in isolated networks
How Zenith works
Drop a file in.
Knowledge comes out.
A fully automatic pipeline turns unstructured files into a searchable, connected knowledge base – without any manual tagging.
01 / CAPTURE
Automatic intake
A watched inbox folder ingests every file: PDFs, Word, Excel, photos, scans. Drop it in – Zenith does the rest, around the clock.
Worker Service · Hot Folder
02 / UNDERSTAND
Multimodal AI vision
A vision-language model reads every page like a human: text, tables, stamps, handwriting. It extracts title, sender, amounts, dates and tax relevance as structured data.
Qwen3-VL 30B · Structured JSON
03 / CONNECT
Knowledge graph
People, companies, projects and places are recognised as entities and linked across documents. Duplicates? Merge them with one click.
Entities · Relations · Merge
04 / ANSWER
Ask anything
Semantic search plus AI chat with source references. Complex questions across document boundaries – answered in seconds, verifiable with one click.
RAG · Vector Search · RerankingNow in prototype development
Your knowledge.
Your next step.
Zenith already makes your archive searchable and answers questions with sources. We are developing local meeting transcription and an agent that connects this knowledge and prepares the next step for your review.
It rebuilds the context
The agent gathers related documents, earlier correspondence, decisions and deadlines from your archive.
It prepares the next step
It prepares reply drafts, checklists, dossiers or reminders with sources. You review the result and approve consequential actions before they run.
You choose the tools
The planned agent will use only tools you enable through MCP (Model Context Protocol). Each tool receives limited permissions; external services stay optional and require explicit activation.
Meetings and documents, connected
Meeting transcripts will join the same searchable knowledge base as your documents, linked to projects and people. The agent can then draw on earlier decisions, contracts and correspondence when preparing its response.
Explore the meeting workflow →Review, approve, trace
The agent is being designed to respect document access rights, show its sources and record its actions in an activity log. Sending, deleting and other consequential actions require your confirmation.
Local meeting transcription
IN DEVELOPMENT
The meeting ends.
The knowledge stays.
A decision should not disappear into an audio file. We are developing a local workflow that turns conversations into knowledge you can find, check and use alongside your documents.
-
Record or import
Start with a local meeting recording or an existing audio file. Transcription is planned to run on your hardware, without a mandatory cloud service.
-
Keep the essentials
Get a searchable transcript, a concise summary, decisions and action items. Review the extracted results before using them.
-
Connect the context
Link the conversation to projects, people and documents. The planned agent can use that context to prepare a follow-up draft, checklist or dossier.
From conversation to project knowledge
Lindenhof project meeting
- Summary
- The team compared two quotes for the renovation. Delivery dates still need to be confirmed.
- Decision
- Request clarification on delivery dates before choosing a supplier.
- Action item
- Anna will request the delivery dates by Friday.
Connected context in your archive
Your next question could be
“What do we need to clarify before placing the order?”
Everything included
Not a document graveyard.
A living knowledge base.
The document core is working today. The Zenith Agent and local meeting transcription are clearly marked as the next capabilities now entering prototype development.
IN DEVELOPMENT
An agent that works with your archive
The planned agent will bring related documents and meeting decisions together, prepare dossiers, draft replies and suggest checklists or reminders. You review the plan and approve consequential actions.
MCP-compatible · Approval gates · Activity log
IN DEVELOPMENT
Private meeting transcription
Local recording or audio import will turn meetings into searchable transcripts, summaries, decisions and action items. Results will connect to projects, people and documents in your archive.
Local speech-to-text · Decisions · Action itemsSemantic search that understands meaning
Search for "all tax-relevant documents from 2025" and find exactly that – even if no file ever contained those words. Vector search compares meaning, not keywords. Optional reranking sharpens the top results.
Vector search directly in SQLAI chat with your documents – incl. attachments
Discuss your document inventory in natural language. Drag a PDF or photo straight into the chat and ask questions about it. Every answer cites its sources – one click opens the original.
Drag & drop · Source referencesAutomatic calendar
Zenith detects appointments and deadlines inside documents and proposes them as calendar entries. You confirm – nothing gets lost.
Deadline detectionEntities & relations
People, companies, projects – automatically recognised and linked across your archive. Merge duplicates manually with full control.
Manual mergeLive processing monitor
Watch in real time which file is being processed, what's queued and how fast the pipeline runs. Full transparency instead of a black box.
Real-time statusStructured extraction – ready for your tax return
Every document is broken down into structured data: title, sender, date, amounts with currency, tax relevance, category. Questions like "all documents for my 2025 tax return" become a single click.
Amounts · Dates · Tax flagsWeb interface – on every device
Zenith runs as a modern web app in your local network. Desktop, tablet or phone: your knowledge base is one browser tab away. No client installation, unlimited users.
Blazor · ResponsiveThe interface
Powerful engine.
Effortless surface.
Enterprise-grade AI shouldn't feel like enterprise software. Zenith's interface stays out of your way – on the desktop and on your phone.
Built for professionals
For every industry.
For every archive.
Wherever critical documents pile up and knowledge lives in filing cabinets – Zenith turns it into instant answers.
Industry & Manufacturing
Machine park knowledge
Manuals, maintenance logs, certificates – ask questions about your entire machine documentation and get instant, precise answers. Offline, right on the shop floor.
Architecture & Engineering
Project & standards archive
Plans, building regulations, correspondence across years of projects – semantically searchable. Deadlines from permits land in the calendar automatically.
Real Estate & Property
Contract intelligence
Leases, utility statements, insurance policies per property – linked as entities. Special termination rights and notice periods surface on time.
Legal & Consulting
Case file research
Client files, opinions, correspondence – professional secrecy demands local processing. Zenith searches thousands of pages without a single byte leaving the office.
Healthcare & Practices
Sensitive records, safe
Reports, referrals and lab results can stay inside the practice network. Local processing supports a data-protection-conscious operating model.
Private & Family Office
Life admin, solved
Insurance, contracts, invoices, warranties – the whole household archive becomes searchable. "All documents for the 2025 tax return" is one question away.
Under the hood
Enterprise AI stack.
Office-sized footprint.
Three specialised AI models orchestrated on one silent mini PC – engineered for 24/7 operation next to your desk.
AMD Ryzen AI Max+ 395 · 128 GB unified RAM
The Minisforum MS-S1 Max delivers workstation-class AI performance: up to 96 GB RAM dedicated to the GPU – enough to run a 30-billion-parameter vision model entirely in memory.
Three specialised models, one team
Qwen3-VL 30B reads and understands documents visually. Qwen3-Embedding 4B builds the semantic index. A dedicated reranker sharpens search results. Each does one job – brilliantly.
Permission-aware agent tools
An MCP-compatible layer will expose only approved capabilities to the agent. Plans, confirmations and activity records keep autonomous assistance bounded and reviewable.
Local speech-to-text pipeline
Meeting audio is planned to be transcribed and structured on the local system, so conversations can become searchable knowledge without a mandatory cloud transcription service.
Vector search inside SQL Server
Embeddings live directly in the database – no separate vector-store infrastructure. Semantic search across thousands of documents in milliseconds, transactionally safe.
Air-gap capable
Zenith needs no internet for everyday operation or AI processing. It can run in an isolated network, reducing the external attack surface and keeping document processing on site.
RAG: answers with receipts
Retrieval-Augmented Generation means the AI answers from your documents, not from vague memory. Every statement links to its source – verifiable in one click.
Modern .NET architecture
Built on .NET 10 and Blazor with a robust worker-service pipeline. Automatic retries, live status, clean data model – engineered for years of unattended operation.
AMD Ryzen AI Max+ 395
128 GB Unified RAM
96 GB GPU Memory
Qwen3-VL 30B
Qwen3-Embedding 4B
Reranker
SQL Vector Search
.NET 10 · Blazor
MCP-compatible tools
Local speech-to-text
Indiegogo · Starts October 1, 2026
Back the software.
Keep the hardware choice.
The campaign funds the software work that turns the prototype into a dependable product: the controlled Zenith Agent, local meeting transcription, installer, security, testing, documentation and onboarding.
Clear separation: Indiegogo rewards are software access and project participation. No computer, GPU, server or other hardware is included.
CROWDFUNDING START01 OCT 2026
Follow Zenith on Indiegogo →
The working document prototype exists. Agent and transcription capabilities are in active development and will be demonstrated as they become ready.
Founding Customer Programme · Market launch October 2026
Shape the launch.
Keep the advantage.
Before we publish final packages, we want to understand your archive, workflow and security requirements. Founding customers receive a personal demo, an individual offer and preferred launch access.
Step 1
Personal demo
See the working product with realistic document workflows.
- ✓Live walkthrough of search, AI chat and knowledge graph
- ✓Architecture and privacy questions answered directly
- ✓No purchase commitment
Founding phase
Step 2
Fit & offer
We define the right hardware, onboarding and backup scope together.
- ✓Needs analysis for archive size, users and network
- ✓Transparent scope, delivery window and total price
- ✓Founding conditions documented in writing
Step 3
Preferred launch access
If the fit is right, founding customers are prioritised for the first installations.
- ✓Early position in the launch schedule
- ✓Direct feedback channel during onboarding
- ✓No subscription and no per-user licence fees
No payment is collected through this website. A reservation only becomes binding after an individual written offer with an agreed scope, price and delivery window.
Frequently asked
Good questions.
Straight answers.
Does Zenith really work without any internet connection?+
Yes – completely. All three AI models, the database and the web interface run on the device itself. Zenith never phones home: no telemetry, no license checks, no cloud fallback. You can operate it in a fully air-gapped network.
Which file formats does Zenith understand?+
PDFs, Word, Excel, PowerPoint, e-mails and images (JPEG, PNG, TIFF – including scans and photographed documents). The vision model reads documents like a human: tables, stamps and even handwriting are captured.
How many users can work with Zenith?+
As many as you like. Zenith runs as a web application in your local network – every device with a browser can access it. There are no per-user licences and no artificial limits.
What happens to my data if the hardware fails?+
Your original files stay in your file system – Zenith works with copies and indexes. Backup requirements are defined during the needs assessment; an external local backup can be included in the individual offer.
Do I need IT expertise to run Zenith?+
No. Zenith is delivered pre-installed and configured for the agreed environment. The exact installation and onboarding scope is documented in your individual offer.
When will Zenith launch – and what does “founding customer” mean?+
The market launch is planned for October 2026. Before then, we invite interested organisations to a personal demo and needs assessment. If both sides see a good fit, you receive an individual written offer with a defined scope, total price and delivery window. Contacting us or attending a demo creates no purchase obligation.
What will the Zenith Agent be allowed to do?+
The agent is in development. Planned workflows include gathering related documents and meeting decisions, preparing dossiers, drafting replies, and suggesting checklists or reminders. Only enabled tools will be available through MCP, with limited permissions and respect for document access rights. Sources and an activity log will make its work reviewable. Sending, deleting and other consequential actions will require confirmation; external services will remain optional.
How will meeting transcription work?+
Meeting transcription is in development. The planned workflow starts with a local recording or an imported audio file. Zenith will transcribe it locally and derive a summary, decisions and action items. Transcripts will become searchable and connect with projects, people and documents, so the agent can include meeting context in its preparation. You review the results before acting on them. Recording participants remains the user's legal responsibility.
Your archive can do more
Stop managing files.
Start moving work forward.
Follow the Indiegogo campaign, see the working foundation and help shape the private AI partner you would trust with your documents and meetings.
Software campaign · No hardware reward · Local-first · Human-controlled agent