Context goes to your answer, not to padding
Long text trimmed with a marker, binary values replaced by their size, empty columns dropped, and only the columns your AI client asked for. Done on your Network Agent before anything leaves your network.
What your Network Agent does for you
Your AI client gets the same rows in a fraction of the space. Your Network Agent trims long text with a marker that says how much was cut, replaces binary values with their size, drops columns that are empty in every row, and sends only the columns the client asked for. The full value is always one call away.
One notes column or one stored file can cost more of an AI client’s context than every other column together. Shaping removes that padding at the source:
| What | What your AI client sees |
|---|---|
| A text cell longer than the limit | The first characters plus a marker such as …[+1,240 chars]. |
| An image, file or other binary value | Its size, such as [blob 4.1 KB], or the column left out. |
| A column empty in every returned row | Left out, and named in the result so the client knows it was there. |
| Columns the client did not ask for | Never sent. The client can ask for exactly the columns it needs, in the order it wants. |
| Decimal numbers | Rounded to the significant digits you choose, or every digit (the default). |
Every trim is counted in the result, with one sentence saying what was left out, so your AI client knows what it did not see and can ask for it.
The plan it needs
| Rencana | This capability | Catatan |
|---|---|---|
| Gratis | Not on this plan | Works as it always has, with standard results. |
| Personal Core, Personal Plus | Termasuk | Part of Context management. |
| Business Core, Business Plus, Business Max | Termasuk | Part of Context management. |
| Enterprise Core, Plus, Max | Termasuk | 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 halaman harga.
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:
Show me the rejected EDI messages, but only the date, partner and status columns.
List the support tickets with the first 100 characters of each description.
Show me the whole description of ticket T-2291, untrimmed.
For people who read tool lists: kolom, max_cell_chars dan full_cells pada qs_jalankan_kueri dan qs_run_saved_query. The result reports cells_truncated, columns_dropped, null_columns_dropped dan blobs_elided, dan shaped_by says the agent did the work. See it on the Tools tab.
What comes back
Show me the latest rejected EDI messages.
Here are the latest three. Long messages are trimmed to 40 characters:
| msg_id | mitra | pesan |
|---|---|---|
| E-88213 | PARTNER-07 | Segment N1 rejected: qualifier ZZ not…[+64 chars] |
| E-88209 | PARTNER-12 | Duplicate control number 000184 for th…[+91 chars] |
| E-88197 | PARTNER-07 | Unknown item code ITEM-9931 in line 4 o…[+52 chars] |
Two of three come from PARTNER-07. Want the full text of E-88213?
Yes, the full one.
E-88213: “Segment N1 rejected: qualifier ZZ not valid for ship-to. Expected 92 or 91. Resend with a corrected N1 segment.”
- …[+N chars]
- The marker at the end of a trimmed cell: how many characters were left out.
- [blob N KB]
- A binary value replaced by its size.
- cells_truncated
- How many cells were trimmed in this result, and in which columns.
- null_columns_dropped
- Columns left out because they were empty in every returned row, named in the result.
- blobs_elided
- How many binary values were replaced by their size.
What it does not help with
- Trimming is not summarising. Your AI client sees the start of the text, not its meaning. Ask for the full value when the wording matters.
- It does not change Excel, Google Sheets, the REST API or Data Viewer. They always get every value in full.
- It does not change Nova unless you switch Apply to Nova pada.
- Rounding with Number precision changes what your AI client sees, not what is stored.
Batasan
- The text cell limit is 50 to 2,000 characters.
- Number precision is off, or 4 to 10 significant digits.
- If none of the columns asked for exists, the call is refused with the result’s real column names.
- Shaping happens on your Network Agent when it is up to date. An older agent still gets shaped results, trimmed on the way back instead, with a note to update the agent.
Pengaturan
Open the MCP page’s Context tab, card What reaches your AI assistant, groups Hasil dan Nova. An admin changes it for the whole organization; changes reach connected AI clients within five minutes.
| Pengaturan | Apa fungsinya? | Bawaan |
|---|---|---|
| Text cell limit | Long text values are cut to this many characters, with a marker showing how much was left out. | 200 characters |
| Binary values | Show size only, atau Leave the column out. | Show size only |
| Drop empty columns | Columns with no value in any returned row are left out. | Pada |
| Number precision | Decimal numbers are rounded to this many significant digits. Mati keeps every digit. | Mati |
| Apply to Nova | When on, Nova’s query results follow the same limits. | Mati |
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. On the Semantic Intelligence page, a column’s AI results menu can also keep that one column always in full or never trimmed; see AI-ready Semantic Intelligence.
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.
| Kode | What the server says | Apa yang harus dilakukan? |
|---|---|---|
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 | None of the columns asked for exists in the result. The message names the result’s real columns. | Ask again with one of the column names it lists. |
Contoh
“What are customers complaining about this week?”
Your AI client reads 300 tickets with each description trimmed, groups them by their opening words, and pulls three full descriptions as examples.
“Which products have no photo?”
The photo column arrives as sizes, not images, so the client counts the empty ones without spending its context on picture data.
Related
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.
Setup trouble?
- Agent shows Offline, or Test connection and Browse are greyed out? Check that the Agent is online → Is the Agent online?

