Answered

Copilot Studio GCC High - 450 KB connector payload limit - alternatives for large file processing

(1) Share

Report

Posted on by MA-27110721-0

Hi everyone,

I'm building a compliance evidence verification solution using Copilot Studio + Power Automate in a GCC High environment. The agent evaluates uploaded documents (PDF, Word, Excel — often 10-100+ pages) against CMMC Level 2 control requirements.

I've confirmed with Microsoft Support that the connector payload limit in GCC is 450 KB (vs 5 MB in commercial). This means I cannot pass large file attachments to my agent via the "Execute Agent and wait" Power Automate action.

My setup:

Multiple customer SharePoint sites (each customer has their own site with evidence documents)
Power Apps triggers Power Automate, which gets the file from the customer's SharePoint site and sends it to the Copilot Studio agent for evaluation
The SharePoint site URL is dynamic — passed at runtime based on which customer is selected

Why "Add as Knowledge Source" doesn't work for me:

I have a multi-tenant/multi-customer system with separate SharePoint sites per customer
Adding all customer sites as static knowledge sources isn't scalable (customers are added/removed regularly)
I need to dynamically point the agent to different sites/documents at runtime

What I'm looking for:

Has anyone found a workaround for the 450 KB limit in GCC High?
Is there a way to pass a SharePoint file URL/link to the agent and have it fetch the file itself at runtime (without pre-indexing as knowledge source)?
Any other patterns for processing large documents with Copilot Studio in government cloud environments?

Thanks in advance for any guidance.

Screenshot 2026-0...

Your file is currently under scan for potential threats. Please wait while we review it for any viruses or malicious content.

Categories:

General topics

I have the same question (0)

All responses (4)

Answers (1)

Sort by

Suggested answer

Ashlesha-MSFT Microsoft Employee on at

Like
a
(0)

Report
Copy link

Link copied!
Hi MA-27110721-0

This is a genuine constraint in GCC High, and your use case is exactly the kind that hits this limit. Here are working workarounds:

1. Chunked Processing (Most scalable)

Split files into <450 KB chunks in Power Automate

Pass chunks sequentially with context (chunk 1 of 5)

Agent evaluates across calls

2. Process First in Power Automate (Cleanest)

Extract/summarize document content before sending to agent

Pass structured data (<450 KB) instead of raw file

3. SharePoint URL Pattern (Dynamic, no pre-indexing)

Pass file URL to agent instead of content

Agent fetches from SharePoint directly

Configure dynamic site URLs at runtime

4. Azure Blob + SAS URL (If available in GCC High)

Store documents in Blob Storage, pass time-limited URLs

Was this reply helpful? Yes No
Verified answer

Assisted by AI

Haque 3,653 on at

Like
a
(1)

Report
Copy link

Link copied!
Hi @MA-27110721-0,

Short answers first:

1. Yes there are workaround for 450KB limit but needs effort.

2. Yes there are ways to pass SP file url/link to the agent - need tricks.

3. Exact common patterns are not there - but you can employ hybrid mechanism

Q-1: Has anyone found a workaround for the 450 KB limit in GCC High?

A: The 450 KB payload limit - we can pass file references (URLs or IDs) no file content having sotred large files in SP or Azure Blob. Extract text or summaries from large documents in Power Automate or Azure Functions, then send smaller chunks [1][2]or summaries within the payload limit. Need to design flows and agents to handle document processing asynchronously, retrieving or processing content in parts rather than all at once.

Q-2: Is there a way to pass a SharePoint file URL/link to the agent and have it fetch the file itself at runtime (without pre-indexing as knowledge source)?

A: Bad news! Copilot agents do not support direct runtime fethcing of files from SP Urls [1] [2] passess as parameters but they work fine as static knowledge source if configured. Dynamic runtime URLs are not supported for direct agent retrieval. In this case - what we can do is pass SP URL/Metadata to agent as input parameters. Then engage Power Automate flow by the agent to fetch and process file content at runtime.

Q-3: Any other patterns for processing large documents with Copilot Studio in government cloud environments?

A: Apart from Copilot Studio, if we desire to have a RAG, there will be generic workflow in terms of what we want. For this particular case there is no stringent pattern out there. As I referred inline - we need to patch up those area to cover up things. But sometimes, when it is huge data ingestion I see and look at this pattern. The answers I discussed above can be lined up in a hybrid mechanism. Which mainly leads to use PowerAutomate or Azure functions to process large documents ( Text extraction, summarization, OCR if needed etc.) by storing processed content in Dataverse, SharePOint or Azure Cognitive Search as indexed knowledge source. Sayin so, we need to split large documents into smaller logical chunks - process chunks incrementally to stay withing payload and limits of the token.

References:

Dynamic Knowledge URLs in Copilot Studio

Use SharePoint content for generative answers

Import sharepoint files with metadata in Azure blob storage

Copy and transform data in Azure Blob Storage

Add knowledge sources to your declarative agent

Add SharePoint as a knowledge source

Handle large messages in workflows using chunking

Parse documents and chunk text as tokenized strings

I am sure some clues I tried to give. If these clues help to resolve the issue brought you by here, please don't forget to check the box Does this answer your question? At the same time, I am pretty sure you have liked the response!

Was this reply helpful? Yes No
MA-27110721-0 30 on at

Like
a
(0)

Report
Copy link

Link copied!

Hi @Ashlesha-MSFT,

thanks for your response.

can you please let me know the process of you 3rd point that is "SharePoint URL Pattern" on how to pass the url, do i just need to pass the url and the agent will fetch it or need to do something else.

and Haque has mentioned that it is not possible.

Was this reply helpful? Yes No
Suggested answer

Haque 3,653 on at

Like
a
(1)

Report
Copy link

Link copied!

Hi @MA-27110721-0,

Thanks for asking. First of all, it's not third point, it is the second point. If you see my short answers (at start of my answer), in point number 2, I mentioned "need tricks". If I mentioned "It is not possible" - that's not a hard stop. If you can get this point "Then engage Power Automate flow by the agent to fetch and process file content at runtime." - we are using intermediary (PA) to store and index content from files.

Please let me know if this suffice.

I am sure some clues I tried to give. If these clues help to resolve the issue brought you by here, please don't forget to check the box Does this answer your question? At the same time, I am pretty sure you have liked the response!

Was this reply helpful? Yes No