query_table returns a limited page by default and reports the full match count next to it, so the model learns how big the result really is without the rows landing in context — and its own description warns against raising that cap without asking you first. Reading the schema first is what makes the OData filters usable, since column names in Table Storage are whatever the writer chose.
A TypeScript server over Azure Table Storage. It connects with a single storage-account connection string and offers three tools: what tables exist, what shape one has, and a filtered query against it.
- Every table in the storage account, optionally narrowed by a name prefix (list_tables)
- The property names and types of a table, so you know what you can filter on before writing a filter (get_table_schema)
- A query with OData filters — equality and inequality, greater and less than, date comparisons, and and, or and not combined (query_table)
- A chosen subset of properties returned instead of whole entities
- The total number of matching rows reported alongside the limited page of results, so you can tell how much you did not see
One Azure Storage connection string, supplied as an environment setting when the client launches the server. Runs from the published npm package or from a local build.
One command plus a key — npx dkmaker-mcp-server-tablestore, then supply credentials