Frequently asked questions about Excalitract

Excalitract reads documents such as invoices, contracts and reports, extracts the values you define in a schema and backs every value with its source location in the original. After review, you export the data as Excel®, JSON or via the REST API into your systems.

No. Excalitract is the data layer in front of it: it delivers reviewed, structured data for your business application, your database or your team to work with. Excalitract does not make business or legal assessments.

Excalitract processes PDF files and PNG and JPEG images, digitally created or scanned. Scanned pages are made machine-readable with state-of-the-art AI-based text recognition (OCR); for digitally created PDFs the existing text layer can be used.

With a schema. You describe the fields you need and provide an example; Excalitract drafts a schema from it and presents any ambiguities as notes for you to confirm one by one. Alternatively, you build fields, tables and layout in the visual editor.

Then the field visibly remains open and is not filled with a zero or a guessed value. Missing required values are flagged and block approval until someone has resolved them.

No. Excalitract has one core for all document types – only the schema changes. Each document type gets its own schema, organised in projects; invoices, contracts and reports run on the same platform.

You upload documents as a folder or ZIP; one run covers up to 200 documents. Duplicate files are detected. Before the start you see a cost estimate, after the run the measured processing time.

Excalitract is designed for documents in German and English.

In the review view, the values are shown next to the original document. Clicking a value jumps to its source location. You confirm, correct or mark values as incorrect; every change is saved as a revision.

For digitally generated PDFs, Excalitract highlights the source location using real coordinates in the document. For scans without a text layer, Excalitract honestly marks the source location as unresolved instead of guessing a box.

That depends on your documents – which is why we don’t quote blanket percentages. In thedemowe work with your document types; on a test set of your documents, business-accepted fields and cost per record can be measured before you decide.

Yes. Follow-up runs process only open or incorrect fields, optionally with a more thorough recognition profile. Values that have already been accepted remain unchanged, and the origin of every value stays traceable.

Yes. Every field carries its origin: extracted by the model, carried over from an earlier run or set by a person. Corrections overwrite nothing – they are stored as a revision.

Yes – measure first, then replace. We compare Excalitract with your current approach on a test set of your own documents. The decision is made on the numbers, followed by a pilot.

Excalitract is operated in data centers in Germany. AI models are accessed exclusively via EU model routes; falling back to other regions is technically ruled out. You can find details underSecurity.

Yes, we provide a data processing agreement (DPA) on request.

Through a native MCP server (Model Context Protocol™) for AI agents, a documented REST API with OpenAPI description and per-tenant API keys, or exports as Excel®, JSON and ZIP. Your team signs in via single sign-on with their Microsoft® account.

Yes. Agents such as Claude®, ChatGPT®, Codex™, Gemini® or Grok® use Excalitract as a tool through the native MCP server. Approvals stay bound to roles, and all access runs through per-tenant API keys and the audit log. More on the home page.

Pricing depends on volume and processing profile. Before every run, Excalitract shows a non-binding cost estimate. We discuss a quote for your case in the demo call.

No. Text recognition and language models are interchangeable via recognition profiles. Schemas, test sets and results belong to you and can be exported at any time.