EN DE
Get in touch

Product

A cockpit, not a toolbox.

CompanyWizard bundles the tools that AI work in a company needs into a single interface — with one login, one privacy pipeline and shared access to your knowledge. Integrated to use, open at the edges.

What is built in

These functions are part of CompanyWizard directly — you do not have to source, connect or align them yourself. Three of them are shown in detail here, the rest in the overview below.

Inquiry chat · RAG

Answers you can back up.

The chat answers from your own knowledge base — every statement with a clickable source reference instead of invented. You pick the model per session, to fit the task.

  • Sources one click away — traceable down into your company documents (RAG)
  • Model choice per task, not global — including a free choice of provider
  • Analyse files directly: PDF, Word, PowerPoint, Excel up to 10 MB
More about the chat in the help →
Inquiry chat: an answer from your own knowledge base, every statement with its source.
Inquiry chat: an answer from the company documents with clickable source references.

Privacy · PII shield

Privacy is built in, not bolted on.

When the PII shield detects personal data in your input, it pseudonymises it inside your own instance — before any language model gets to see it. A notice shows you transparently what was protected.

  • Pseudonymisation runs inside your own instance
  • A notice appears whenever personal data is detected
  • Applies to document export as well, not just the chat
How the PII shield works →
The PII shield marks personal data before the request reaches the model.
The PII shield marks detected personal data before the model call.

Knowledge base · Vaults

Your knowledge — searchable and reusable.

A curated, searchable Markdown collection the assistant reads from and writes to. Good chat results go back into the knowledge base with a single click.

  • Multiple vaults, role-based — everyone sees only what they may
  • Hybrid search: full text and semantic combined
  • Adopt chat answers — knowledge grows as you work
More about the knowledge base →
A solid answer travels back into the knowledge base with one click.
A chat answer is saved into the knowledge base.

The complete cockpit at a glance

Seven built-in functions, each with its specifics — all details visible right away, nothing hiding behind tabs. Individual cards can be collapsed with a click.

01 Inquiry chat Chat grounded in your knowledge base — with source references and privacy pre-processing.
  • Answers stream live; model selectable per session
  • Source references (RAG) — every statement backed by your vaults
  • PII shield pseudonymises sensitive input inside your instance
  • File analysis: PDF, Office, text/code up to 10 MB
  • Agent delegation — complex tasks go to specialist agents
  • Visual answers — diagrams & charts from structured data
  • Personal memory ("remember this") + incognito switch
Open in the help →
02 Image generation Images from text — with knowledge-grounded prompt optimisation, a gallery and provenance.
  • Two-stage — first prompt optimisation, then image generation
  • Knowledge-grounded optimisation (RAG) — "Based on …" sources, role-based
  • Frame of your choice — provider, model, aspect ratio, image size
  • Reference images — up to 4 templates (PNG/JPEG/WEBP, max. 10 MB each), model-dependent
  • Consistent style (admin) — a company-wide image optimisation prompt governs the optimisation
  • Refine produces targeted variants
  • Gallery — save, search and delete images with metadata; move them into knowledge articles
  • Transparency — C2PA Content Credentials + labeling under EU AI Act Art. 50(2)
Open in the help →
03 Prompt library Proven instructions kept reusable, instead of lost on sticky notes.
  • Create, edit and delete your own prompts — stored per user
  • Company prompts — company-wide, admin-maintained, visible to all (read-only)
  • Placeholders {{name}} — a dialog asks for the values on use
  • Categories: general, writing, analysis, code, translation, custom
  • Apply directly in the chat; available in web and desktop
Open in the help →
04 Skills Predefined capabilities and workflows — extensible via ZIP, without a restart.
  • Searchable catalogue with the full instruction view
  • Activation per session (chat) or per agent ("Skills" tab)
  • The AI reads the instructions on demand via tool calls
  • Reload without a restart (rescan); admins upload new skills as ZIP
  • Origin chip shows: shared catalogue or local source (locally overridable, desktop)
Open in the help →
05 Knowledge base A curated, searchable Markdown collection — read and write.
  • Multiple vaults, role-based — you see only what you may
  • Search combines full text + semantic (sqlite-vec/Qdrant/BM25)
  • Read & edit including frontmatter; switchable editor (source/WYSIWYG)
  • Embedded images & PDFs are rendered
  • Overview/landing pages per vault and folder (Übersicht.md, index, README)
  • Import single files or a ZIP; adopt chat answers
Open in the help →
06 Messages Server Share chat results with colleagues directly, without a detour through another program.
  • 1:1 and group chats with a persistent history
  • Conversation list with search and unread markers
  • Share AI answers and knowledge articles
  • 1:1 dialogues with agents — the bot replies in the conversation
  • A short cooldown per agent guards against loops
  • Note: licensed server operation
Open in the help →
07 Export to PDF & Word Every article and every chat result in your own corporate layout.
  • Formats: Word (.docx) and PDF
  • AI preparation — title/subtitle suggestions, text cleared of filler
  • Metadata: title, subtitle, author, date, version, confidentiality
  • Your corporate template with bookmarks is filled automatically
  • Respects the PII shield (cleanup before the AI preparation)
Open in the help →

Connected tools

On top come tools that connect the cockpit to your systems — one tool, one learning curve, one place for security and governance.

  • Git (beta) — Working with repositories and code.
  • SQL (beta) — Questions to your databases in plain language.
  • Mail — Reading and composing messages.
  • Calendar (beta) — Appointments and planning as context — read-only.
  • Contacts — Search and read address books — read-only over CardDAV.

Open at the edges

For CompanyWizard, integration does not mean lock-in. The cockpit has doors to the outside, in three directions:

  • Extensible — embed your own tools via MCP servers (beta), with free choice of LLM provider.
  • Reachable — the knowledge base is itself reachable via an MCP server, for other programs and cloud agents too.
  • Portable — your content lives in Obsidian-compatible Markdown. No lock-in that depends on a vendor’s promises.

The honest "self-built stack vs. CompanyWizard" comparison, including the limits of the approach, lives in the Open architecture concept.

Calendar integration (beta)

Ask your calendar: on request, the assistant reads your events as context — "What’s on this week?", "Am I free Thursday at 3 pm?". Read-only across CalDAV and Microsoft 365, each person connects their own account, GDPR-friendly.

More on calendar integration · Help: connect a calendar and use it in chat.

Contacts

Who was that again? On request the assistant searches your address books — "What is Maria Schmidt's email address?", "All contacts at Muster GmbH". Read-only over CardDAV, through the same account as the calendar (Microsoft 365 provides no contacts).

Because contacts are third-party data, they run through the PII shield by default — unlike your own appointments.

Help: connect an account and use it in chat.

Your concrete use case? Get in touch.