Service Description

This is a convenience translation. In case of doubt, the authoritative German version prevails.

Provider: evanto media GmbH, Brunnstraße 25, 93053 Regensburg, Germany ("evanto").

As of: 26 September 2026 · Version: 1.1

This service description is Annex 1 to the License Terms for CompanyWizard and forms part of the contract. It describes the scope of functions, limits, system requirements, support and updates per edition, as well as Managed Operation and the Managed AI add-on. In case of conflict the License Terms prevail. It applies in the version published at the time the contract is concluded. Response, backup and availability figures in this service description are target values, not warranted characteristics.

1. General

(1) CompanyWizard is an AI chat and knowledge platform for businesses, delivered as a desktop application (Windows, macOS) and as a web application (Docker container).

(2) Operating modes: single workstation (desktop, autonomous), desktop server mode (several desktop clients on one instance) and web operation (browser clients on one server instance). Availability of the modes depends on the edition (section 2).

(3) AI models are provided by the Customer (own accounts with model providers or local models via Ollama / LM Studio) unless the Managed AI add-on is booked (section 8). Web search always requires the Customer's own account. Providers the Customer integrates itself remain the Customer's responsibility, also under Managed Operation and with Managed AI.

(4) System requirements: Desktop: the current and the previous major version of Windows and macOS. Web: Docker-capable Linux host as per the deployment documentation, with outbound HTTPS to the license service, the catalog service and the connected providers. Optional, operated by the Customer: Qdrant (vector database) and Presidio (AI-assisted PII detection) as Docker services.

(5) License check and offline grace: activation by e-mail and company name (double opt-in); periodic online check-in. Without a connection the software keeps working for 14 days in single-workstation mode and 30 days in server/web mode.

(6) Capacity limits (all editions, default values): chat attachments up to 10 MB per file (text and source code files, PDF, Word, PowerPoint, Excel, images); files for document import up to 50 MB, ZIP archives for import up to 100 MB; skill uploads as ZIP archive up to 100 MB; input per chat turn up to 60,000 characters. The administrator can adjust individual limits in the configuration. Further limits are set out in the documentation.

2. Editions overview

Business hours are Monday to Friday, 9 am to 5 pm (CET/CEST), excluding nationwide public holidays in Germany and public holidays in Bavaria.

3. Functions in all editions

Unless expressly limited to an edition below, all editions include:

• AI chat with streaming, session history, file attachments (section 1 (6)), model selection, export of answers as Word, PDF, PowerPoint and Excel.
• Knowledge base (RAG) with vaults, file and folder browser, Markdown editor, full-text/BM25 search; semantic search where a vector database is provided; pattern-based basic PII detection.
• Document import (Word, Excel, PDF, HTML) via the import page, with conversion (offline or by text recognition via an own account or Managed AI) and release into the knowledge base.
• Skills (prompt templates and capabilities), prompt library, image studio (with an own image account or via Managed AI), Personal Memory.
• External tools (MCP) and connection of several own AI model providers (BYO) as well as local models.
• Manual AI agents including dry run.
• Web search (with an own account).
• Research mode over the knowledge base.
• Backup and restore (desktop: encrypted portable backup; vault export in both modes).

4. Free (Solo)

(1) Use free of charge for one named workstation in autonomous desktop operation.

(2) Functions as per section 3. Excluded are all functions that require multi-user operation (central user management, team messaging, server-side scheduled agents, LDAP / directory sign-in).

(3) The Customer provides the infrastructure (vector database, PII engine, AI models) itself. Without a vector database, local search (BM25 or sqlite-vec) is used.

(4) No entitlement to support or specific updates.

5. Pro

(1) Use against a fee per booked named seat. Pro enables multi-user operation in desktop server mode (several desktop clients on one desktop instance of the Customer), with central user management and enforcement of the seat limit. Web operation is reserved for the Site edition.

(2) Functions as per section 3, plus: team messaging in desktop server mode; PII pseudonymisation with audit log; automated agents and skills; own MCP server for the knowledge base; LDAP / Active Directory sign-in; SharePoint and Confluence import; continuous synchronisation of imported sources with the knowledge base; AI-assisted re-ranking of search results (reranking); workflows; model comparison; SQL database as agent tool; Git tools for working folders; calendar tools (Microsoft 365 / CalDAV).

(3) A named seat may be reassigned to another person at most once per calendar month (seat change).

(4) Updates: all versions released during the term. Support: e-mail on business days with the response target of answering a request by the next business day; no warranted response time.

6. Site

(1) Use against a fee for one company instance at one location with up to 100 named users; more users or further locations by separate agreement.

(2) Functions as Pro, plus web operation and continuously running, server-side scheduled agents in web operation.

(3) Operation as a dedicated instance in the Customer's environment or by evanto under Managed Operation (section 7).

(4) Additionally: directory / single sign-on integration on request; onboarding as one online session of up to two hours for administrators and users, including a recording; named contact. Further training and consulting are services as per section 9.

(5) Updates as Pro. Support: prioritised handling with the response target of 8 business hours, the contact can be reached by phone on business days; no warranted response time.

7. Managed Operation (operated by evanto)

(1) Under Managed Operation evanto operates the Customer's instance in evanto's infrastructure (data centre R-KOM GmbH, Regensburg, EU). Managed Operation is an operating form of the Site edition and can be booked with it. The License Terms (version 1.3) call it "Managed Infrastructure". It is independent of Managed AI and includes no AI model access. The Data Processing Agreement applies in addition.

(2) Tiers:

(3) Maintenance windows are announced in text form at least two business days in advance and lie outside business hours. Security-critical emergency maintenance may take place without notice; in that case evanto informs the Customer without undue delay afterwards. Exceeding the seat or data limits by agreement.

(4) Without tier L the Customer provides AI-assisted PII detection itself or uses the pattern-based basic detection.

(5) A warranted availability with a price-reduction consequence can be agreed for tier L in an individual contract; without such an agreement paragraph 2 applies as a target value.

8. Managed AI add-on

(1) evanto provides access to AI models through the aggregator Cortecs GmbH, Vienna and bills usage against a fixed monthly budget. Bookable with Pro (per seat) and Site (as a package). Managed AI is independent of Managed Operation and includes no operation of the instance. The Data Processing Agreement applies in addition; evanto engages Cortecs as a sub-processor on the basis of the Cortecs.ai data processing agreement.

(2) Included: chat and text generation, embeddings for the knowledge base, AI-assisted re-ranking of search results (reranking), text recognition (OCR) on document import and image generation; text recognition and image generation insofar as the aggregator provides a suitable model for the booked variant. Use of all included functions is charged against the monthly budget. For text recognition and image generation the Customer may additionally integrate its own accounts; their costs are borne by the Customer directly vis-à-vis the provider and are not charged against the monthly budget. Not included: web search; this requires the Customer's own account.

(3) Quality levels: Fast, Balanced, Thorough plus an auto mode choosing the level per request; evanto maintains the specific models per level and publishes them in the software (admin area). From 80 % budget consumption the auto mode selects the most economical level; a level chosen by the user is kept. When the budget is exhausted, requests via Managed AI are declined until the first of the next month; own accounts and local models remain usable. There is no additional billing.

(4) Variants and budget:

(5) Data processing: zero data retention at the aggregator, no use for training. evanto provides the current list of model providers used via Cortecs, with region, on request.

(6) Consumption and budget status can be viewed in the admin area of the software; the Customer can additionally set its own cost and usage limits per user and model there.

9. Not part of the editions, Managed Operation or Managed AI

• costs of the Customer's own model, OCR, image and web search accounts (BYO) and local hardware.
• integration of the Customer's own source systems beyond the standard importers (file, Word, Excel, PDF, HTML, SharePoint, Confluence, Git working folders).
• custom development, consulting and training beyond the scope named in section 6; the general terms and conditions of evanto media GmbH for services apply to those.
• operation of the instance without booked Managed Operation; AI model access without booked Managed AI.