Copilot Studio / Work IQ Graph API Discovery Failure: 20-Item Pagination Cap Blocks Processing of Comprehensive SharePoint Directory Structures

Yes, learn.microsoft.com (specifically the Microsoft Q&A section under the microsoft-copilot-studio tag) is a great place to post this for community support and visibility.

💡 Pro-Tip on Channels: If your client has an active enterprise agreement, the absolute best place to file an official bug ticket with an SLA is through the Microsoft 365 Admin Center (admin.microsoft.com) or the Power Platform Admin Center (admin.powerplatform.microsoft.com) under Help & Support.

Here is a professionally structured headline and technical description you can copy-paste into your forum post or support ticket to get the engineering team's attention.

Headline

Copilot Studio / Work IQ Graph API Discovery Failure: 20-Item Pagination Cap Blocks Processing of Comprehensive SharePoint Directory Structures

Description

Environment Context:

Product: Microsoft Copilot Studio (Agent Builder)

Features Enabled: Generative Answers (Grounding on SharePoint Knowledge Source), Work IQ.

Target Environment: SharePoint Online Document Library featuring a flat structure containing over 141 subdirectories housing customer invoice PDFs.

Issue Summary:

When utilizing a Copilot Studio agent configured to ground knowledge from a specific SharePoint parent folder, the agent fails to discover, index, or interact with files situated past the initial 20–50 directories.

The agent throws immediate truncation warnings during processing or outright states it cannot paginate past the first payload slice it receives via the underlying Microsoft Graph API. It yields a static list of the first 20 results and completely ignores the remaining 121 directories.

Furthermore, attempts to forcefully call or trigger queries against the broader dataset often trigger transient gateway errors, including API Error: 400 status code (no body).

Observed Behavior & Technical Bottlenecks:

Lack of Dynamic Graph API Pagination Loop: The standard Copilot Studio orchestrator is failing to parse or loop through the @odata.nextLink tokens returned by Microsoft Graph, halting discovery on the first index page.

Knowledge Boundary Thresholds: The engine treats the 141 individual paths as separate entities and caps processing at the documented 20-source / 50-folder ingestion limit, despite being pointed at a unified parent directory.

Work IQ Alignment: Even with Work IQ enabled for deep semantic mapping, the system inherits the payload cap constraints and fails to execute wide-scale structural lookups.

Impact:

This completely blocks the delivery of external, client-facing enterprise agents intended for wide-scale document discovery, analytical aggregation (e.g., parsing total spend across invoices), and directory mapping.

Requested Resolution:

Does Microsoft plan to provide native automatic server-side pagination looping for deep SharePoint hierarchies within Copilot Studio / Work IQ, or is the architecture strictly bounded to require external preprocessing frameworks (like flattening arrays into Dataverse)?Yes, learn.microsoft.com (specifically the Microsoft Q&A section under the microsoft-copilot-studio tag) is a great place to post this for community support and visibility.

💡 Pro-Tip on Channels: If your client has an active enterprise agreement, the absolute best place to file an official bug ticket with an SLA is through the Microsoft 365 Admin Center (admin.microsoft.com) or the Power Platform Admin Center (admin.powerplatform.microsoft.com) under Help & Support.

Here is a professionally structured headline and technical description you can copy-paste into your forum post or support ticket to get the engineering team's attention.

Headline

Copilot Studio / Work IQ Graph API Discovery Failure: 20-Item Pagination Cap Blocks Processing of Comprehensive SharePoint Directory Structures

Description

Environment Context:

Product: Microsoft Copilot Studio (Agent Builder)

Features Enabled: Generative Answers (Grounding on SharePoint Knowledge Source), Work IQ.

Target Environment: SharePoint Online Document Library featuring a flat structure containing over 141 subdirectories housing customer invoice PDFs.

Issue Summary:

When utilizing a Copilot Studio agent configured to ground knowledge from a specific SharePoint parent folder, the agent fails to discover, index, or interact with files situated past the initial 20–50 directories.

The agent throws immediate truncation warnings during processing or outright states it cannot paginate past the first payload slice it receives via the underlying Microsoft Graph API. It yields a static list of the first 20 results and completely ignores the remaining 121 directories.

Observed Behavior & Technical Bottlenecks:

Lack of Dynamic Graph API Pagination Loop: The standard Copilot Studio orchestrator is failing to parse or loop through the @odata.nextLink tokens returned by Microsoft Graph, halting discovery on the first index page.

Knowledge Boundary Thresholds: The engine treats the 141 individual paths as separate entities and caps processing at the documented 20-source / 50-folder ingestion limit, despite being pointed at a unified parent directory.

Work IQ Alignment: Even with Work IQ enabled for deep semantic mapping, the system inherits the payload cap constraints and fails to execute wide-scale structural lookups.

Impact:

This completely blocks the delivery of external, client-facing enterprise agents intended for wide-scale document discovery, analytical aggregation (e.g., parsing total spend across invoices), and directory mapping.

Requested Resolution:

Does Microsoft plan to provide native automatic server-side pagination looping for deep SharePoint hierarchies within Copilot Studio / Work IQ, or is the architecture strictly bounded to require external preprocessing frameworks (like flattening arrays into Dataverse)?

I tried to execute queries using copilot chat and it complains similar issue citing limitation with Graph API limitation.