# Crossary

> Crossary is AI-assisted source-to-target data mapping for integration and migration teams. Upload source and target specification documents (Excel, PDF, CSV, JSON, XML, XSD, SQL, or YAML) and it proposes a mapping for every target field — each suggested row carrying a self-rated confidence that is a triage cue, not a guarantee, and, where the mapping rests on a rule rather than a plain name match, the spec line behind it (rendered verbatim, so a weak match shows itself). When nothing fits, it abstains instead of guessing: the row becomes an honest gap with a clarifying question, because a wrong mapping is worse than an honest gap.

A human reviews every row. Validation is deterministic — zero AI. Validation checks structure and key semantics (type compatibility, source-field references, cardinality) — it is not a full correctness sign-off. Export is an attested .xlsx workbook — a stamped record of who reviewed it and when, not a cryptographic signature — that can be edited in Excel, Google Sheets, or Numbers and re-imported without losing a reviewer note.

Crossary maps the spec, not the data, so in most cases production records and PII never need to be uploaded. The free tier includes 3 one-time integration credits with no card; reviewing, validation, export, and round-trip re-import never cost credits on any plan.

Built for integration and interface engineers, implementation and onboarding consultants, data and migration engineers, and EDI analysts — anyone who hand-builds the mapping sheet someone else implements.

## Pages

- [Home](https://www.crossary.com/): what Crossary does and the five-stage pipeline (artifacts → fields → mapping → validation → export)
- [Free source-to-target mapping template](https://www.crossary.com/templates/source-to-target-mapping-template): a free, ungated .xlsx — the same Mapping Specification layout the product exports, sample-filled
- [AI data mapping](https://www.crossary.com/ai-data-mapping): how AI source-to-target mapping works here — evidence, self-rated confidence, honest abstention
- [Guides](https://www.crossary.com/guides): the guide index — the mapping document, the process that produces one, and where spreadsheets break
- [What is a source-to-target mapping (STTM) document?](https://www.crossary.com/guides/source-to-target-mapping-document): the definition, what a complete STTM contains, and one real populated row
- [How to create a data mapping document](https://www.crossary.com/guides/how-to-create-a-data-mapping-document): the columns, the six steps, and a fully populated example row from a real run
- [Source-to-target mapping in Excel](https://www.crossary.com/guides/source-to-target-mapping-in-excel): why Excel is the right review medium, the five ways workbooks break, and the round-trip rules that survive them
- [LLM data mapping (engineering notes)](https://www.crossary.com/guides/llm-data-mapping): why confidence alone misleads, why evidence must be verbatim, why abstention is a first-class output
- [Worked examples](https://www.crossary.com/examples): complete outputs of real runs over public vendor specs, published unedited
- [QuickBooks Online → NetSuite worked example](https://www.crossary.com/examples/quickbooks-to-netsuite-customer-mapping): 29 proposed customer mappings, 110 honest gaps, 9 blocker questions — unreviewed AI output, not a finished workbook
- [QuickBooks Online → Business Central worked example](https://www.crossary.com/examples/quickbooks-to-business-central-customer-mapping): 18 source-bound mappings, 64 fields explained as the target's own machinery, 75 honest gaps, 7 implementer questions
- [Salesforce → HubSpot worked example](https://www.crossary.com/examples/salesforce-to-hubspot-contact-mapping): 22 source-bound contact mappings, 116 properties explained as HubSpot's own machinery, 33 honest gaps, 4 migration questions
- [Guided sample](https://www.crossary.com/sample): a worked HubSpot → Salesforce contact mapping, end to end — free, no upload, no card
- [Pricing](https://www.crossary.com/pricing): free tier and paid plans; AI runs cost credits, reviewing never does
- [Security](https://www.crossary.com/security): how uploaded specs are stored and handled
- [Support](https://www.crossary.com/support): how to reach us