Data Freshness and Changed-Since via MCP

Analyse your data with your AI client · Data freshness and changed-since

Know how current the data is, and read only what changed

Ask when a table was last updated. Re-run a question and get only the rows that are newer than last time, with a marker to carry forward.

Personal Core and up Runs on your Network Agent Solo lectura, siempre

¿Tienes algún problema? Pregúntanos. Abrir Ayuda En Consultas de flujos (menú superior derecho → Soporte) y envíanos un mensaje. Personas reales lo leen, y «No sé qué me pide esta pantalla» es una pregunta perfectamente válida.

What your Network Agent does for you

Your AI client can tell you how fresh a table is before it answers from it, and re-run a question for only the rows that changed. Your Network Agent reads the newest value of the table’s timestamp column and applies the “since last time” filter at the source, so a repeat question returns only what is new.

Preguntar	Lo que sucede
“How current is the orders table?”	Your agent reads the newest value of the table’s timestamp column and reports it, with the column it used and how long the read took. No rows.
“What is new since this morning?”	The query runs with a since-filter on the timestamp column and returns only newer rows.
The next time	The answer carries a watermark: the newest value it read. Your AI client passes it back to read only what changed after it.

The timestamp column is chosen for you: the connector’s default changed-since column if one is set, otherwise the table’s first date or timestamp column. Your AI client can also name one, including a number column such as an increasing id.

The plan it needs

Plan	This capability	Notas
Gratis	Not on this plan	Works as it always has, with standard results.
Personal Core, Personal Plus	Incluido	Part of Result analysis.
Business Core, Business Plus, Business Max	Incluido	Part of Result analysis.
Enterprise Core, Plus, Max	Incluido	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 precios.

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 moreSolo lectura

When was the shipments table last updated?

qs_data_freshness

Show me the orders that came in since 9 this morning.

qs_run_query · since_column · since_value

Anything new since you last checked?

qs_run_query · since_value=<watermark>

Groundwork done on your Network Agent, inside your network registrado en la pestaña Actividad de MCP

For people who read tool lists: qs_data_freshness (connector, table, optional column); since_value y since_column en qs_run_query y qs_run_consulta_guardada. The result’s diagnostics carry the new watermark. See freshness on the Tools tab, y changed-since.

What comes back

Claude

ChatGPT

Cursor

Código Claude+10 moreSolo lectura

Is the shipments table up to date? Then show me what shipped since 9.

qs_data_freshness qs_run_query · since_value

The newest shipment is from 10:42 today (column enviado_en), so the table is current. Since 09:00, 3 new shipments:

shipment	carrier	enviado_en
SH-50291	CARRIER-B	09:14
SH-50292	CARRIER-A	10:05
SH-50293	CARRIER-B	10:42

Next time I will ask only for rows after 10:42.

Groundwork done on your Network Agent, inside your network watermark 10:42 carried forward

Newest value: The maximum of the timestamp column: how current the table is.
Column and source: Which column was used and why: named on the call, the connector’s default, or picked from the table.
watermark: The newest value read. Passed back next time, it returns only rows after it.
Notas: One sentence of caution when needed: a date-only watermark can miss rows added later the same day, and when a row cap or budget cut the answer the watermark is withheld so no rows are skipped.

What it does not help with

It sees new and updated rows only through the timestamp column. A table whose update column is not maintained will not show its updates.
Deleted rows do not appear: they have no new timestamp. To compare two states of a table, see Result diff.
A table with no date, timestamp or number column cannot be read this way.
It does not schedule anything. For a repeating check, use an alert.

Límites

Freshness waits 60 seconds by default, at most 300. A table without an index on the column is scanned whole; an index makes the read instant.
File-backed and API sources cannot apply a watermark, and SQLite connectors do not support it by design. Filter in the WHERE clause there instead.
Some statements cannot be wrapped with the filter, for example one with a TOP or LIMIT clause or a query hint.

Configuración

The default changed-since column is set per connector, on the connector’s settings page. Without one, the table’s first date or timestamp column is used.

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	¿Qué hacer?
`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_CHANGED_SINCE_NO_COLUMN`	“no timestamp column is known for this query, so since_value has nothing to compare against; pass since_column (a date or timestamp column of the result), or set the connector’s default changed-since column on its settings page.”	Tell your AI client which column to use, or set the default on the connector.
`MCP_CHANGED_SINCE_UNSUPPORTED`	The connector’s source cannot apply a watermark at all.	Filter in the WHERE clause instead.
`MCP_CHANGED_SINCE_UNWRAPPABLE`	The source can, but this statement’s shape cannot be wrapped.	Remove the TOP or LIMIT clause, or filter in the WHERE clause.
`MCP_AGENT_VERSION_TOO_OLD`	“Update the Network Agent to report data freshness; this agent cannot answer qs_data_freshness yet, so read the maximum of the timestamp column with qs_run_query until it is updated.”	Agents update themselves. Check the agent under Datos → Agentes de datos.
`MCP_AGENT_NOT_CONNECTED`	The Network Agent that owns the connector is not connected right now.	Abrir Datos → Agentes de datos and start the agent on its computer.

Ejemplos

“Before you build the weekly summary, is the data in?”

Your AI client checks freshness first and tells you the invoices table stopped at Thursday 18:00, so Friday is missing.

“Keep me posted on new returns.”

Each time you ask, the client passes the last watermark and gets only the returns logged since, usually a handful of rows.

¿Sigues atascado? Abrir Ayuda En Query Streams, indíquenos en qué paso se encuentra y qué observa. Le ayudaremos a conectarse.

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?