Profile a Million Rows in One MCP Call

Analyse your data with your AI client · Profile a result

A million rows, summed up in one page of text

Your Network Agent reads the whole result and sends your AI client the shape of the data: counts, ranges, averages, top values, empty rates and a fair sample. The rows never leave your network.

Personal Core and up Runs on your Network Agent Somente leitura, sempre

Teve alguma dúvida? É só perguntar. Aberto Suporte Em Fluxos de Consulta (menu no canto superior direito → Suporte) e envie-nos uma mensagem. Pessoas reais leem as mensagens, e "Não sei o que esta tela está pedindo" é uma pergunta perfeitamente válida.

What your Network Agent does for you

Your AI client can reason about a million rows from one call. Your Network Agent streams through the result once inside your network and sends back per-column statistics and a small sample: about 2,400 tokens for a profile of a million rows. On a table Semantic Intelligence has analysed, the profile can come straight from that analysis without reading the table at all.

Without a profile, your AI client reasons about the first thousand rows and guesses about the rest. With one, it starts from the whole result. Per column you get:

Statistic	What it tells your AI client
Count and empty values	How many values are filled in, and how many are missing.
Distinct values	An estimate of how many different values there are.
Minimum and maximum	The range, for numbers, dates and text.
Average and spread	For numbers.
Text lengths	Shortest and longest value, for text.
Most common values	The top 10 by default, for columns with a manageable number of different values.
Meaning	The column’s semantic type and description, when Semantic Intelligence knows it.
Sample rows	20 whole rows by default.

Situação	Before	With a profile
“Analyse this year’s West-region orders.”	The client saw 1,000 of 200,000 rows.	One call covers all 200,000.
“What values does status take?”	The client guesses from a page of rows.	Every status with its count.
“Is anything missing?”	Unknown.	The empty rate of every column.

The plan it needs

Plano	This capability	Notas
Gratuito	Not on this plan	Works as it always has, with standard results.
Personal Core, Personal Plus	Incluído	Part of Result analysis.
Business Core, Business Plus, Business Max	Incluído	Part of Result analysis.
Enterprise Core, Plus, Max	Incluído	Part of Result analysis.

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 página de preços.

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

Cursor

Código Claude+10 moreSomente leitura

Profile this year’s West-region orders.

qs_run_query · mode=profile

Profile the daily inventory table, but only the first million rows.

qs_run_query · mode=profile · max_rows

Give me today’s figures, not the stored ones.

qs_run_query · mode=profile · prefer_live

Groundwork done on your Network Agent, inside your network Acessei a guia Atividade do MCP.

For people who read tool lists: mode: profile sobre qs_executar_consulta e qs_executar_consulta_salva, com top_k (5 to 25), sample_rows (1 to 100), max_rows e prefer_live. See it on the Tools tab.

What comes back

Claude

ChatGPT

Cursor

Código Claude+10 moreSomente leitura

Profile the daily inventory table.

qs_run_query · mode=profile

Profiled the first 1,000,000 of about 60 million rows, as the table is large:

column	filled	distinct	range / top values
item_code	100%	≈ 8,400	ITEM-0001 … ITEM-9998
bin	14%	≈ 5,900	86% empty
estado	100%	4	A 71% · H 18% · C 9% · X 2%
qty	100%	≈ 5,200	−40 … 12,000, average 38

Bin is empty on 86% of rows, so a filter on it would miss most stock. Want the sample rows?

Groundwork done on your Network Agent, inside your network about 2,400 tokens for 1,000,000 rows

rows_scanned: How many rows were read, with the estimated size of the whole result when the scan stopped early.
capped_by: Why the scan stopped: the row limit, the size limit, or the end of the result.
profile_source: live when your agent read the rows, semantic_intelligence when the table’s Semantic Intelligence analysis answered it, with when that analysis ran.
When it is too big: If a profile would exceed the call’s budget, the sample rows go first, then the most common values, and a warning says which were left out.

What it does not help with

A profile reads every row it covers, so a slow query is still slow. A shape first tells your AI client how long a full scan would take.
Most common values are only listed for columns with a manageable number of different values. Others are marked high-cardinality and rely on the samples.
It is not an export. To see rows, ask for rows.
For a quick look at one table without writing a query, tabela_de_perfil_qs answers from Semantic Intelligence on every plan, Free included.

Limites

Up to 1,000,000 rows by default; your AI client can ask for up to 10,000,000. On large tables, use a filter or a row limit.
A profile returns no rows, so your plan’s rows-per-query limit does not apply to it.
A profile does not combine with holding a result. Hold the rows in a separate call if you also want follow-ups.
A plain read of one table that Semantic Intelligence has analysed can be answered from that analysis. Queries with joins or filters are always read live.

Configurações

Your AI client sets the number of most common values, sample rows and rows to read on each call. Your organization’s defaults are 10 most common values and 20 sample rows. The What your AI client sees preview on the MCP page’s Context tab runs a saved query through your settings.

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.

Código	What the server says	O que fazer
`MCP_FEATURE_NOT_INCLUDED`	“Your plan does not include result analysis; profiles, data-quality reports and follow-up questions on a held result are available on Personal Core.”	Ask for rows instead, or move up to Personal Core.
`MCP_AGENT_VERSION_TOO_OLD`	“Update the Network Agent to profile results at the source; this agent cannot answer mode=profile yet, so ask for rows or a shape until it is updated.”	Agents update themselves. Check the agent under Dados → Agentes de Dados.
`MCP_QUERY_TIMEOUT`	“Stopped waiting after {n} s; the Network Agent had {read} {rows} rows at about {rate} rows per second. Profile with max_rows ≤ {suggested} or narrow with a WHERE clause.”	Ask for a profile with the suggested row limit or a filter.
`MCP_AGENT_NOT_CONNECTED`	The Network Agent that owns the connector is not connected right now.	Aberto Dados → Agentes de Dados and start the agent on its computer.

Exemplos

“What does a typical order look like this year?”

Your AI client profiles 200,000 orders and answers with the average value, the busiest month, the four order statuses with their shares and the share of orders with no discount code.

“What values does the carrier column take?”

The table has Semantic Intelligence, so the profile comes back from the analysis with a note of when it ran. “Use today’s data” reads it live.

Ainda com dificuldades? Aberto Suporte Em Fluxos de Consulta, informe-nos em qual etapa você está e o que você vê. Nós o conectaremos.

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?