Sketch nodes, relationships and properties. Catch the duplicates and orphans while it is still a diagram. Then export the schema as Cypher, GraphML, Mermaid or a data dictionary your team can actually read.
One 200 KB HTML file. No account, no install, no dependencies — it works offline, straight from your downloads folder.
The diagram is a view of a real schema — labels, property keys, cardinality, uniqueness — so the tool can tell you when the model contradicts itself.
Give every node its labels and property keys, with required, unique and indexed flags. Relationships carry direction, type and cardinality.
Duplicate node names, duplicate relationship types between the same pair, orphaned nodes and missing keys are flagged in a running list — not at export time.
Drag, pin and group nodes. Changing a node's shape never changes the space it occupies, so a settled diagram never reshuffles itself underneath you.
Eleven formats, because the schema has more than one audience: the database, the diagram tool, the docs, and the person who just wants a spreadsheet.
Cypher — constraints, indexes and parameterised CREATE / MERGE statements.
GraphML opens in yEd, Gephi and Cytoscape. Mermaid renders in GitHub, Notion and Confluence.
Three CSV shapes: a one-row-per-relationship review sheet, a detailed sheet with cardinality and notes, and a data dictionary with one row per property.
Bring an existing model in and carry on. Auto-detects the format; replace the current model or merge into it.
The same tool. The difference is only where the model lives and whether anyone else can see it.
| Single file | Web app | |
|---|---|---|
| Account | None | Required |
| Where the model lives | Your browser's local storage | Your account, synced |
| Works offline | Yes — it is one HTML file | No |
| Modelling, validation, exports | All of it | All of it |
| Version snapshots | Yes, pruned to fit local storage | Yes, unlimited |
| Sharing and roles | — | Owner, editor, commenter, viewer |
| Comments in context | — | Yes |
| Change log | — | Yes |
| Price | Free | Free |
Data models get argued over before they get built. The web edition is shaped around that argument rather than around two people typing at once.
Threads attach to a node or a relationship, so "this should be a relationship, not a property" lands on the thing it is about.
Entries are generated from the same diff that writes the data, so the history cannot drift from what was actually saved.
Label a version before a risky change and restore it whole. Replacing a model by import snapshots it first.