EN DE
Book a 30-minute call

Product

CompanyWizard. The AI platform that stays in your house.

Chat, knowledge base, document import, image generation, and automated agents under one interface. With automatic personal-data pseudonymisation before any external handover, and a swappable language model — Anthropic Claude, OpenAI GPT, Google Gemini, AWS Bedrock, Vertex AI, local via Ollama or LM Studio. One instance per company, web or desktop — the single seat is free.

Request a demo See the architecture
CW · BETA

Why a platform of your own?

Your people want to use AI — and they already do, just outside the company. Texts go into public chats, documents go to external translators. What is convenient is also a double loss: sensitive data leaves the building, and your own company knowledge stays invisible to the AI.

What CompanyWizard does

Six building blocks, one surface.

  • AI chat as the central workspace

    Streaming answers in real time, attachable files, swappable model per task. Answers with clickable source references, export to Word or PDF in your corporate design. RAG, PII protection, and tools work with every model — including local ones via Ollama or LM Studio.

  • Knowledge base as your second brain

    Open Markdown files, Obsidian-compatible — the "second brain" as Andrej Karpathy describes it. Word, PDF, Excel, and images become semantically searchable: built-in SQLite-Vec vector database, optionally Qdrant. Separate vaults per department, four confidentiality levels, role-based access.

  • From original to searchable knowledge

    Word & PDF (offline or via Mistral OCR), HTML & web URLs (tracker-free), Confluence, SharePoint, X threads, and email mailboxes (IMAP/POP3/Exchange). Every import passes through human review.

  • Agents for recurring work

    Input source, personality, tools with read/write markers, cron scheduling. A dry-run shows the plan before any file is touched. Live timeline per real run, permanent audit log.

  • Image generation with prompt optimisation

    Images are created right in the interface — built in, no extra tool. Automatic prompt optimisation refines your input before the image model sees it: better results without prompt tinkering.

  • Team chat for research results

    In the multi-seat installation: a built-in chat for your team. Share LLM results seamlessly instead of copy-pasting — and chat directly with active agents to check on progress.

Privacy that does not get in the way

Personal data is automatically pseudonymised.

Before any handover to a language model — chat, tool result, agent run — content passes through a two-stage personal-data filter: first pattern-based, then AI-driven on the basis of Microsoft Presidio, in German and English. Presidio runs as your own Docker installation or hosted by evanto. Names become [PERSON_1]; the answer is re-identified before output. The shield applies regardless of the chosen model — including local ones. A status indicator in the header shows at any time whether the shield is active; the audit log secures compliance.

Document PII PII masking LLM Language model Re-identification
1 · Document 2 · PII masking 3 · Language model 4 · Re-identification

Architecture & technology

A dedicated instance, swappable building blocks — no vendor lock-in.

  • No LLM lock-in, no knowledge lock-in: switch the language model by configuration — cloud provider or local model — and your knowledge base stays openly accessible via MCP, for example straight from Claude Desktop. If a provider changes its terms, you change the configuration, not the platform.
  • One dedicated instance per customer — no shared cloud, no multi-tenancy risk.
  • Native desktop app for macOS and Windows or web (Docker container) — same interface, maximum independence.
  • Built-in SQLite-Vec vector database for RAG; optionally Qdrant — as your own Docker installation or hosted by evanto.
  • Swappable language model: Anthropic Claude, OpenAI GPT, Google Gemini, AWS Bedrock, Vertex AI — or local via Ollama and LM Studio.
  • RAG, PII, and tool pipeline identical for all models — including local ones.
  • PII workflow based on Microsoft Presidio — as your own Docker installation or hosted by evanto.
  • Extensible via Skills (ZIP upload, hot-reloadable) and external tools through Model Context Protocol (MCP).
  • Role-based access control with four confidentiality levels, personal vaults per user.
  • Stack: .NET 10, Blazor WebAssembly, SignalR, Microsoft Agent Framework, Microsoft.Extensions.AI, Aspire.

Editions & pricing

Start free, grow with your team.

Single seat

Free Desktop app for macOS and Windows

  • Full feature set: chat, knowledge base, import, agents, image generation.
  • Knowledge base Obsidian-compatible — open Markdown files.
  • Built-in SQLite-Vec vector database for RAG.
  • Your own API key or local models via Ollama / LM Studio.
  • PII workflow optional — requires a Presidio installation (Docker or hosted by evanto).

Download available soon — until then on request.

Get in touch

Team / Company

from 14.00 EUR per active user / month, plus VAT

  • Dedicated instance — web (Docker) or desktop, one shared knowledge base.
  • Built-in team chat: share LLM results directly, chat with active agents.
  • Role-based access, four confidentiality levels, audit log.
  • Qdrant and Presidio as your own Docker installation or hosted by evanto.
  • Setup of a dedicated instance: time and material (one-off).

Definition of "active user": monthly active users (MAU) — only people who actually use the system.

Request a demo

Market comparison

Comparable platforms (Glean, Microsoft 365 Copilot, ChatGPT Enterprise) charge 25–70 EUR per user per month. We start deliberately low — you pay for the solution, not for the vendor's marketing budget.

What is not included

  • Model costs: LLM API consumption runs on your own API key — or does not apply with local models via Ollama / LM Studio.
  • Your own AI-model hardware (for local models on-premise).
  • Integration with your own source systems beyond the six standard importers.

All prices plus statutory VAT.

Currently in beta — looking for pilot customers

Six months of free use for the first ten beta partners.

CompanyWizard is currently in beta. We are looking for ten mid-sized companies (Mittelstand) who will use the product for six months free of charge — in exchange for their requirements and feedback. The product matures in real operations rather than in the lab.

All post-beta prices — including the free single-seat edition — are in the pricing overview.

Apply as a beta partner Request a demo

What we deliberately do not automate

Five places where the person decides.

  1. Final approval of imports

    Conversion, deduplication, and suggested metadata are automatic; the move into the knowledge base is explicitly human.

  2. Live runs of agents

    Dry-run plan first, then a deliberate decision to run for real.

  3. Confidentiality classification

    The system suggests, the person decides.

  4. Binding legal advice

    The AI references and links; interpretation stays with the responsible person.

  5. Multi-tenancy

    Deliberately not. One dedicated instance per company. Other tenants' data is technically unreachable because it does not live in the same system.

Good fit for

Six industries we currently serve particularly well.

Click an industry to open its detail page — or, if not yet built, the industries hub.

30-minute demo with your own requirements

See CompanyWizard live in action.

We set up the demo instance for your use case — either with synthetic example data or, under NDA, with a slice of your own content. Or start on your own: the single-seat desktop edition for macOS and Windows is free — download available soon.

Request a demo Book a 30-minute call