Configure advanced agent responses with custom knowledge sources
Unlock the complete study guide + 1,040 practice questions across 16 full exams.
Bundled into the existing Designing and Building Integrated AI Agent Solutions in Copilot Studio premium course — no separate purchase.
14-day money-back guarantee — no questions asked.
Included in this chapter:
- Where knowledge is configured, and what a source actually is
- The catalog of source types, and what each one costs
- Two SharePoint options in one dialog
- Bounding a SharePoint source with Advanced settings filters
- Authentication is part of the source, not a wrapper around it
- Public website sources and the Bing boundary
- Scoping a source at runtime with a variable
- Custom data: supplying the grounding yourself
- Selection metadata: getting the orchestrator to pick the right source
- Reading a silent no-result
- How these choices show up in questions
Comparing the source types you configure most, along the axes that vary between them
| Axis | Public website | SharePoint (full integration) | SharePoint (file upload) | Dataverse |
|---|---|---|---|---|
| Content path: where the content lives | On the public web, reached through Grounding with Bing Search | Stays in SharePoint | Copied into Dataverse from SharePoint | In the configured Dataverse environment |
| Content path: freshness | Whatever Bing has indexed | Real time, reflects the latest available content | Synchronized every four to six hours | Queried at request time |
| Identity: authentication | None | Agent user's Microsoft Entra ID authentication | User credentials checked live against the source at query time | Agent user's Microsoft Entra ID authentication |
| Retrieval scope you configure | One URL, up to two levels of depth, variable allowed | Site URLs plus Advanced settings filters on Title, Author, Modified by, Modified on | The specific files and folders you selected | Tables in the environment |
| Published input count in generative mode | 25 websites | 25 URLs | Counts against the agent's knowledge objects | Unlimited |
| Pick it when | The content is public, indexed, and you want no ingestion | The library changes often, or you need filters and sensitivity labels | You want vector semantic search over a fixed set of files and can accept the refresh delay | The grounding data is business records already in Dataverse |
Decision tree
Cheat sheet
Unlock with Premium — includes all practice exams and the complete study guide.
Also tested in
References
- Knowledge sources summary - Microsoft Copilot Studio
- Add a generative answers node - Microsoft Copilot Studio
- Orchestrate agent behavior with generative AI - Microsoft Copilot Studio
- Unstructured data as a knowledge source - Microsoft Copilot Studio
- Add SharePoint as a knowledge source - Microsoft Copilot Studio
- Quotas and limits - Microsoft Copilot Studio
- Generative answers pointing to SharePoint sources don't return results - Copilot Studio
- Add a public website as a knowledge source - Microsoft Copilot Studio
- Use environment variables in Power Platform solutions - Power Apps
- Use a custom data source for generative answers nodes - Microsoft Copilot Studio
- Make HTTP requests - Microsoft Copilot Studio