Schema That Fits: Large Databases via MCP

Analyse your data with your AI client · Schema that fits

A database of thousands of tables, navigated by business area

Instead of one schema listing too large for any AI client, your Network Agent and Query Streams serve the business areas first, then the tables in one area, then full detail for the tables your AI client names.

Personal Core and up Runs on your Network Agent Sempre in sola lettura

Hai bisogno di aiuto? Chiedi pure a noi. Apri Supporto Accedi a Query Streams (menu in alto a destra → Supporto) e inviaci un messaggio. Persone reali lo leggono, e "Non so cosa mi stia chiedendo questa schermata" è una domanda assolutamente legittima.

What your Network Agent does for you

Your AI client finds its way around a database of any size the way an analyst does: business areas first, then the tables in one area, then the full detail of the tables it actually needs. On a very large ERP the first answer is a short list of areas, where the whole-database listing would not fit in any conversation.

Without this, “what tables are there?” is one call that returns every table and every column. On a small database that is fine. On a real ERP the answer alone is bigger than the AI client can read, so the call fails and the client starts guessing table names. With Schema that fits, the same question gets a short, layered answer:

Level	What your AI client gets	When
Business areas	Each area with its table count and three example tables.	First answer on a database with more than 300 tables and more than one business area.
Tavoli	One lean row per table: name, estimated row count, business area. No columns.	First answer above 40 tables, and whenever the client asks for one business area or one schema.
Compact	One line per table with column names and types.	When the client asks for it. Coding clients also get it first on a small database.
Completo	The complete tree, 25 tables per page.	Small databases, or when the client asks for it.

Situazione	Before	With Schema that fits
“What is in our ERP?”	The listing is too large for one call and fails.	The business areas with table counts, in one short answer.
“Where are the invoices?”	The client guesses a name and the query errors.	The client lists the Finance tables with row estimates and picks the right one.
“Total last month’s credit notes.”	A wrong column, then a retry.	Full detail for the two tables it named, so the first query is more often right.

The plan it needs

Piano	This capability	Note
Gratuito	Not on this plan	Works as it always has, with standard results.
Personal Core, Personal Plus	Incluso	Part of Context management.
Business Core, Business Plus, Business Max	Incluso	Part of Context management.
Enterprise Core, Plus, Max	Incluso	Part of Context management.

Why it is on paid plans: your Network Agent does the heavy lifting next to your database, so the rows stay in your network and your AI client spends its attention on the thinking. That groundwork is what the paid plans add.

The Free plan works as it always has. Paid plans add the Analytic toolset. Free is a plan you keep: MCP, the REST API, federated queries and alerts, with the same results as before. Plan names and prices are on the pagina dei prezzi.

How to ask

You do not call tools yourself. Ask in plain words and your AI client picks the tool. Three asks that bring this capability in:

Claude

ChatGPT

Cursore

Codice Claude+10 moreSola lettura

What business areas does the ERP database have, and how many tables are in each?

qs_get_connector_schema · detail=domains

List the tables in the Finance area with their row counts.

qs_get_connector_schema · domain=Finance

Show me everything about the invoice header table before you query it.

qs_get_table_schema

Groundwork done on your Network Agent, inside your network accesso effettuato nella scheda Attività MCP

For people who read tool lists: qs_get_connector_schema takes dettaglio (domains, tables, compact, pieno), domain, schema, pagina e page_size. qs_get_table_schema is unchanged and still returns one table in full. See it on the Tools tab.

What comes back

Claude

ChatGPT

Cursore

Codice Claude+10 moreSola lettura

We just connected the ERP. What is in it?

qs_get_connector_schema schema_intelligence

It is a large database: well over a thousand tables, grouped into business areas. The biggest:

Business area	Tavoli	Esempi
Vendite	419	sales_order, order_line, price_list
Finanza	212	gl_journal, ar_invoice, ap_payment
Inventario	188	stock_level, stock_move, warehouse
Purchasing	140	po_header, po_line, supplier
Produzione	96	work_order, bom_line, routing

More areas hold the remaining tables. Which area should I open?

Groundwork done on your Network Agent, inside your network a short answer for a very large database

Business area: Comes from Semantic Intelligence, which groups tables into areas such as Sales or Finance. Tables without one are listed under (nessuno).
Tavoli: How many tables the area holds. The client asks for one area next and gets its table list.
Row estimate: On the table list: roughly how many rows each table holds. The client uses it to avoid scanning a huge table by accident.
Misurare: Every schema answer reports roughly how many tokens it cost, and says what the next level down would add.

What it does not help with

It does not write descriptions that do not exist. Business areas and table descriptions come from Semantic Intelligence; a database where it has not run gets the table list without areas.
It does not change the per-table answer. qs_get_table_schema returns the same full detail as before.
A compact column list for one very large area can still be big. The client only asks for it when you do, and the table list is usually enough.
Nothing here writes to your database or to your stored schema.

Limiti

The full level is paged: 25 tables a page by default, at most 100. An oversized page is shortened, and the first full answer on a small but very wide database is swapped for a lighter level.
A filter that matches nothing returns one sentence naming the business areas or schemas that do exist.
Business-area navigation works on databases whose schema Query Streams has captured. A database outside the connector’s capture selection is read live, needs the key’s run-queries scope, and supports a compact listing for SQL Server, MySQL and PostgreSQL.

Impostazioni

Open the MCP page’s Context tab, card What reaches your AI assistant, group Schema. An admin changes it for the whole organization; changes reach connected AI clients within five minutes.

Collocamento	Cosa fa	Predefinito
Schema view for large databases	What your AI client gets first for a large database: Table list by business area, Compact column list o Dettagli completi.	Table list by business area
Large database threshold	Databases with more tables than this start with the view above; smaller ones get full detail.	40 tables

When your AI client says no

When the server refuses, your AI client receives a code and one sentence, and usually relays it. Quoted sentences are the server’s own words; the parts in italics stand for the ids a real message names.

Codice	What the server says	Cosa fare
`MCP_OUTPUT_TOO_LARGE` (Free)	On Free the whole-database call behaves as it always has, and on a very large database it is too big. The answer then carries one sentence: “This database is too large to list in one call, so navigate it by business area on Personal Core.”	Ask for one table by name, which works on every plan, or move up to Personal Core.
`MCP_FEATURE_NOT_INCLUDED`	“Your plan does not include context management; schema by business area, shaped results and a size budget are available on Personal Core.”	Your AI client carries on without the input and gives you the plain result. To keep the capability, move up to Personal Core.
`MCP_INVALID_INPUT`	A `dettaglio` value that is not one of the four levels.	Ask for domains, tables, compact or full.
`MCP_DATABASE_NON_ACQUISITO`	The database is outside the connector’s capture selection, so there is no stored schema to navigate.	Add it to the capture selection on the connector, then run Semantic Intelligence for business areas.

Esempi

“Where do customer credit limits live?”

Your AI client reads the business areas, opens Sales and Finance, spots the customer master table, reads its full detail and answers with the column and a sample value. Four small calls instead of one failed listing.

“Which Inventory tables have more than 10 million rows?”

The client asks for the Inventory table list, sorts by the row estimate and names the big ones, so you know which to filter before querying.

Ancora bloccato? Apri Supporto Accedi a Query Streams e comunicaci a quale passaggio ti trovi e cosa visualizzi. Ti aiuteremo a connetterti.

Facts source

Setup trouble?

Agent shows Offline, or Test connection and Browse are greyed out? Check that the Agent is online → Is the Agent online?