What a DATEV export carries
The German entity’s books reach Control.dev as a DATEV Kontenblatt, the account ledger export, through your AI client over MCP. The client inspects the file and explains its structure, maps the local accounts (SKR04 in the documented example) to the group chart of accounts, and saves the postings once you confirm them: one file upload, the faster route for a large ledger, or batch by batch, up to 100 rows each. It then reads the saved entries back and compares the account and monthly totals with the source.
- Postings with booking text, document numbers and counteraccounts
- Opening balances for the balance sheet
- Mappings from local accounts to group accounts
What an AI client gets from DATEV alone today
DATEV’s Developer Portal lists Online APIs, for exchanging and processing data online, and Desktop APIs, an interface available only in the local network and not over the internet1. In July 2026 DATEV announced its Datenservice Export Rechnungswesen, with which software can retrieve bookkeeping data directly for the first time2. Its API, accounting:dataexchange, provides accounting data from the last archiving in DATEV Rechnungswesen and authenticates through OAuth 2.0 with OpenID Connect3.
The portal lists accounting:dataexchange as a paid API1. Each data service goes live through an onboarding with DATEV: the first costs EUR 1,500 plus VAT, including four hours of consulting, and DATEV’s advisers review the solution before the app reaches production4. DATEV’s own AI assistant is currently available only to tax advisers and auditors5.
What Control.dev’s MCP adds above DATEV
Control.dev takes the German entity’s DATEV export in through your AI client, puts it beside the group’s other ledgers and gives the whole group to the same AI client over MCP.
- Every entity on one group chart Each entity’s accounts are mapped to one group chart of accounts, and the eliminations, currency translation and CTA are done before the client reads. The eliminations sit in a column of their own.
- Statements first, not a raw export The client starts from the statements and the group structure, not a raw export of the general ledger. It finds what needs attention there and traces it to the entries.
- Every figure opens to its entries Every actual opens to the transactions behind it, and every row stays connected to its source system and legal entity.
- Your role decides what it sees Over MCP the client gets at most the access your role allows in Control.dev, and a read-only session rejects write tools.
- A plan with a reason on every cell The client can also draft budget values over MCP, and every value it writes carries its reason. The budget version keeps its change history, so a later variance analysis compares the actuals with the assumptions.
- Other MCP servers in the same conversation The client can use its other MCP servers in the same conversation, a CRM for example, and bring that data into a plan or an analysis next to the group’s actuals.
How each AI client connects is on the Control.dev MCP server, every tool it exposes on MCP tools and reusable instructions on finance skills, every integration and what comes in from it on the integrations page, and how the group view is built on multi-entity consolidation.
Already connected Claude to DATEV yourself? Control.dev vs building it yourself with Claude compares what you build and keep alive, what the client reads, who may change what, and what it costs.
DATEV MCP in three steps
Our team sets up the integration with you. Most companies have live data in under two hours from the first call, a larger group takes days, the hardest a few weeks, and there is no setup fee.
- 01Export the Kontenblatt from DATEV Export the entity’s Kontenblatt from DATEV. With your AI client connected to Control.dev (the next step), drop the file into the conversation: the client explains it and proposes the account mappings, and nothing is saved until you have reviewed the accounts and totals and confirmed the import in your AI client. The steps are in the DATEV upload guide.
- 02Copy the server URL and the client ID into your client In Control.dev, open Data outputs and then MCP Server. Add the server URL and the client ID from the MCP Server page to Claude, ChatGPT or Codex as a remote MCP server, then sign in with your Control.dev login. If the page shows no client ID yet, ask us: registering your client is part of setup.
- 03Ask something you already know One company, one month, one figure. Check it against the same view in Control.dev, then ask the bigger questions.
Questions
What does Control.dev cost for a group with an entity on DATEV?
The DATEV entity comes in as an export through your AI client, and the group’s other entities connect through their own integrations, one integration being one connected accounting system, such as one entity’s ledger. Control.dev publishes its monthly prices, excluding VAT: EUR 249 for one entity and one integration, EUR 499 for consolidation with three integrations and EUR 999 with five, where a Stripe connection counts as one integration, with no implementation fee, no multi-year contract and a notice period from one month. Larger groups are priced on request: book a demo.
Does Control.dev write to DATEV?
No. No connection to DATEV exists, so nothing goes back to it. The export comes in through your AI client over MCP and is saved in Control.dev once you confirm the import in your AI client.
Where does the AI process our data?
Control.dev’s own AI features process data in the EU, and Control.dev’s data is hosted on Google Cloud in Hamina, Finland, encrypted in transit and at rest. What your AI client reads is processed under your agreement with its provider, so where that data goes is your decision.
How does a correction in DATEV reach the group view?
Through a new export, not a sync. Check the corrected Kontenblatt’s accounts and totals against the source as before. For a file upload, submit the same upload again with the corrected file: it replaces that upload’s rows, while a new upload would add to them. For rows written in batches, agree with your finance team how to correct or reverse them before writing new ones. Then let the client read the saved entries back and compare them.
What do DATEV’s data services do, and when do you use both?
They are for software that integrates with DATEV directly: a data service DATEV announced in July 2026 lets software retrieve bookkeeping data directly2, and each data service goes live through an onboarding with DATEV4. Use them when your own software should read DATEV. Use Control.dev for the group: the German entity’s DATEV books beside the other ledgers on one group chart of accounts, with the eliminations done. Your AI client reads both in one conversation: the DATEV export you drop in and the group in Control.dev.