Can AI create a real XLSX workbook?

Yes. OfficeMaker creates native Microsoft Excel (.xlsx) files from a brief or structured JSON, using the live Excel schema.

Can OfficeMaker query Excel before sending data to an LLM?

Yes. A workbook can be filtered or aggregated deterministically so the model receives the relevant result set rather than the complete workbook.

Can generated spreadsheets contain formulas and charts?

Yes where those structures are supported by the current Excel schema and included in the payload. Review generated formulas before relying on them.

Can an AI agent create Excel through MCP?

Yes. An MCP client can fetch the Excel schema and create a native XLSX through OfficeMaker. Available operations depend on the selected free or authenticated document-service surface.

Excel

AI Excel generator for native workbooks

OfficeMaker creates native Microsoft Excel (.xlsx) workbooks from a brief or schema-led JSON. For data-heavy work, it can query and filter existing workbooks deterministically first, so the language model reasons over the relevant rows rather than the entire file.

Direct answer

Yes. OfficeMaker creates a native, editable Excel (.xlsx) workbook. The published example, generated on 22 October 2026, has Assumptions and Forecast sheets and 12 formulas, including cross-sheet references and SUM. It has no chart. Formulas in the payload omit the leading equals sign. Querying an existing workbook before the model sees it is a separate step.

A workbook with formulas

The published Excel example was generated on 22 October 2026. Assumptions holds the inputs. Forecast calculates three months with 12 formulas, including Assumptions!B2 and SUM. Formula text in the payload omits the leading equals sign. That file has no chart, because the payload did not define one. The starting revenue of 42000 is a demonstration input.

Download the Excel forecast

What you get

The result is a real .xlsx workbook, not a screenshot or a markdown table. Where the live Excel schema supports them, the payload can define sheets, tables, values, formulas, charts, number formats and workbook structure.

The workbook remains editable in Excel. That matters for finance and operations teams that need to inspect formulas, refresh inputs, reuse the file and keep a conventional Office artifact after the AI step.

Claim. OfficeMaker creates native XLSX files from schema-led structured payloads.

Mechanism. An agent or application reads the live Excel schema, produces structured JSON, validates the payload, and the OfficeMaker document middleware writes the workbook.

Evidence. Document generation API

Limitation. Only features present in the current Excel schema should be assumed. Generated formulas and financial outputs still require review.

Filter first, reason second

A large spreadsheet is usually a poor prompt. The stronger pattern is source workbook → deterministic query/filter/aggregation → relevant result set → LLM reasoning → Office output.

This reduces context noise and can reduce model-token consumption because irrelevant rows never enter the model context. It also keeps calculations and filtering in deterministic code instead of asking a language model to approximate spreadsheet logic.

Query or aggregate the workbook before model reasoning.
Pass the smallest useful structured result set to the model.
Use the LLM for explanation, classification, narrative and document structure.
Keep arithmetic, filtering and file construction deterministic where possible.

Typical Excel use cases

Budgets, forecasts and scenario workbooks
Recurring operational and management reports
Quotes, payout and assumption-driven models
Board reporting workbooks and supporting charts
Agent-generated workbooks from API, MCP, CRM or form data

How the architecture works

OfficeMaker separates model reasoning from Office-file implementation: schema → structured JSON → validation → OfficeMaker middleware → native XLSX.

JSON and HTML-like structured representations are heavily represented in modern LLM training data. OfficeMaker deliberately gives the model a familiar structured representation rather than asking it to reason directly about XLSX internals. That is an architectural design choice, not a guarantee that every prompt will outperform every alternative.

Where OfficeMaker fits

OfficeMaker is the document-execution layer, not a replacement for Excel, Power Automate, Zapier, Make or n8n. Those systems can orchestrate the business process; OfficeMaker can be the step that queries, creates or updates the workbook.

For repeatable reports, use the automation pattern rather than rebuilding the workbook from a fresh prompt every month.

Limitations and review

Generated workbooks are not automatically audited financial models. Review formulas, assumptions, ranges and number formats before relying on them for financial decisions. The public free MCP surface is narrower than authenticated paid document-service capabilities and does not expose every existing-file operation.

Create an account · Developer resources

Questions

Can AI create a real XLSX workbook?: Yes. OfficeMaker creates native Microsoft Excel (.xlsx) files from a brief or structured JSON, using the live Excel schema.
Can OfficeMaker query Excel before sending data to an LLM?: Yes. A workbook can be filtered or aggregated deterministically so the model receives the relevant result set rather than the complete workbook.
Can generated spreadsheets contain formulas and charts?: Yes where those structures are supported by the current Excel schema and included in the payload. Review generated formulas before relying on them.
Can an AI agent create Excel through MCP?: Yes. An MCP client can fetch the Excel schema and create a native XLSX through OfficeMaker. Available operations depend on the selected free or authenticated document-service surface.