What you need
- DATEV DMS or Document Storage with the DATEVconnect Dokumentenmanagement licence and an online, up-to-date Klardaten connector.
- Document search enabled for your organization and instance, plus access to the instance, DATEV profile and Document Management module.
- For use in an AI assistant: Klardaten MCP access and a connected MCP-compatible application, such as Langdock, ChatGPT or Claude.
Enable search and ask your first question
- As an administrator of your organization, open the required instance in the customer portal and select the “Document search” tab.
- Select “Enable document search”. Setup and indexing continue in the background even after you close the page.
- Select “Refresh status” to check progress. Once “Search available” appears, you can start searching; more documents may still be processing.
- Open your connected AI application and ask a specific question about document content. If its search tool is still missing, refresh the available tools or reconnect the integration.
- Check the cited sources. If “Setup needs attention” appears, read the displayed error and use “Retry setup” after resolving it.
Try these questions in your AI assistant
- Find letters in which the tax office requests further evidence for business entertainment expenses. Show the relevant passages.
- Find agreements about private use of a company car.
- Find documents explaining a provision for litigation costs.
Example questions: results depend on your indexed documents and the connected profile’s permissions.
Then ask: “Summarize the documents requested in the letter you found and cite your sources.”
How does search work?
Klardaten creates a local search index with a vector database on the DATEV connector host. Search combines keywords with semantic similarity. Text extraction, OCR and search processing run locally on that machine.
Retrieved passages and associated document information are sent through Klardaten to your connected AI application. Further processing is governed by your agreements and settings with that provider.
Scope and common questions
Which documents are prepared?
Initial setup saves a fixed cutoff three years before that point in time. Documents created in DATEV after this cutoff are considered. This uses the DATEV creation timestamp, not the invoice date or tax year printed on the document. The cutoff does not move forward every day; it is retained across later starts. Older documents are therefore not automatically included.
Which file formats are supported?
Content search processes PDFs with extractable text, and supported scans and images through OCR (automatic text recognition), including PNG, JPEG and TIFF. TXT, Markdown, CSV, JSON, XML and HTML are also processed. Native content extraction for Word, Excel, PowerPoint and Outlook files is currently unavailable. A filename match therefore does not mean that the file body was searched. The quality and availability of extracted text depend on the source document.
When does new content become searchable?
Initial preparation can take time depending on document volume and machine capacity. New setups index from 20:00 to 06:00 Europe/Berlin on weekdays and throughout weekends. Existing settings are retained. Prepared documents remain searchable outside indexing hours. Changes in DATEV are processed subsequently, so search is not a guaranteed current view of the entire archive. Portal status is refreshed manually.
Can I configure indexing hours and resource use?
Activation applies to your organization on the selected instance. Shared indexing hours and processing-worker counts require an additional management permission. These settings affect all organizations using the same instance. Worker counts limit concurrency rather than directly limiting CPU or memory consumption.
Which sources do I receive?
Search returns relevant passages with filenames and document information; page numbers are included where available from processing. Ask your AI assistant to cite these sources. Search retrieves relevant documents, while the AI assistant writes the answer. An empty result list does not prove that no matching document exists. Search does not replace a complete file or deadline review.
Which permissions apply?
Before results are returned, document permissions are checked through DATEVconnect using the connected DATEV profile. People sharing a profile share its permissions. Use individual profiles when people should have access to different documents.
Where is data stored, and what happens when I disable search?
Extracted text, document information and the search index are stored on the connector host. Matching passages are sent through Klardaten to the connected application during a search. Disabling search revokes your organization’s search access but does not delete the local index. Indexing can continue if another organization still uses the same instance. Removing the index is a separate management operation.
If document search is unavailable for your organization or setup continues to fail, contact Klardaten support. support@klardaten.com