Token Budget: Results Sized for Your AI Client, Never Cut Silently

Analyse your data with your AI client · Token budget

Every result reports its ukuran. Nothing is cut silently.

Your AI client can set a size budget for a result. Whole rows are kept until it is spent, and whatever did not fit is reported with a link to the complete result.

Personal Core and up Runs on your Network Agent Hanya baca, selalu

What your Network Agent does for you

Your AI client always knows how much of its context a result will use, and never gets a result that was quietly cut. The budget is resolved for every call: the size the client asks for, otherwise your organization’s setting, otherwise a default for the client in use. Whole rows are kept until the budget is spent, and the rest is reported with a link to the complete result.

An AI client reads every result it is handed. When a result is too large, most clients truncate it, drop it, or stop and ask you to narrow the question, and nobody tells you which rows were lost. With a token budget:

• Every result says how big it is, in tokens, the unit an AI client’s context is measured in.
• Your AI client can set a budget for a call. Whole rows are kept in order until it is spent. A row is never cut in half, and at least one whole row always comes back.
• What did not fit is reported: how many rows were left out, one sentence on how to get them, and a link that opens the complete result in Query Builder.
• The budget follows a fixed order: the size the client asks for on the call; otherwise your organization’s Result size if an admin set one; otherwise a default for the AI client calling (see Client-aware defaults).

The plan it needs

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.

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 Kursor Kode Claude+10 moreHanya baca

Show me last month’s orders, but keep the answer under 8,000 tokens.

qs_run_query · max_tokens=8000

How much of your context did that last result use?

(reads the size from the last result)

Give me what fits and a link to the full result for the rest.

qs_jalankan_kueri

Groundwork done on your Network Agent, inside your network masuk pada tab Aktivitas MCP

For people who read tool lists: max_tokens pada qs_jalankan_kueri dan qs_run_saved_query (at least 2,000; a smaller value is raised to the floor). The result reports approx_tokens, budget_tokens, budget_source dan budget_rows_dropped in its diagnostics. See it on the Tools tab.

What comes back

Claude ChatGPT Kursor Kode Claude+10 moreHanya baca

List every open order line with its item and quantity. Keep it under 8,000 tokens.

qs_run_query · max_tokens=8000

The first rows of 2,000 fit the 8,000 tokens you asked for. The first three:

order_no	item	qty	due
SO-10418	ITEM-2201	40	2026-10-21
SO-10418	ITEM-2310	12	2026-10-21
SO-10422	ITEM-0457	300	2026-10-24

The rest were left out, not lost: Open the complete result in Query Builder. Want me to group them by due week instead, so it all fits?

Groundwork done on your Network Agent, inside your network within the 8,000 tokens asked for

approx_tokens

Roughly how many tokens the result costs in your AI client’s context.

budget_tokens

The budget that applied to this call.

budget_source

Where the budget came from: input (the call itself), org (your organization’s Result size), or client_default (the default for your AI client).

budget_rows_dropped

Whole rows left out to stay inside the budget. When it is above zero the answer also carries a link to the complete result.

What it does not help with

• It does not make the query cheaper for your database. The query runs the same; only what reaches your AI client is sized.
• It does not summarise the rows it leaves out. To learn about all of them, ask for a profil.
• It does not change what Excel, Google Sheets, the REST API or Data Viewer receive. They always get the full result.

Batasan

• The smallest budget is 2,000 tokens, so a typo can never return an empty result.
• Your plan’s rows-per-query limit applies first, then the budget.
• Tokens are an estimate from the size of the result. Your AI client may count slightly differently.
• Results count against your data allowance the same way as every other client.

Pengaturan

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

A budget your AI client sets on a call beats the organization setting, and the organization setting beats the per-client default. Each key can differ: when you create or edit a key on the MCP page, its Result shaping section sets Inherit (follow the organization), Mati atau Strict (a request cannot loosen the organization’s settings), plus the key’s own Result size. Itu What your AI client sees preview on the same tab runs a saved query through your saved settings and shows the size the client would get.

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.

Contoh

Related

• Column and cell shaping
• Client-aware defaults
• Shape first
• Analyse your data with your AI client: every capability
• Connect Query Streams to your AI client via MCP: setup, permissions and troubleshooting

Facts source

Facts on this page: the Query Streams MCP server and Network Agent as of 19 October 2026, with the server’s own refusal sentences quoted word for word. Table and column names in the examples are made up.