Compare · SharePoint agents
SharePoint agents alternative for cross-app cited search without the 20-source limit
Agents in SharePoint are a genuinely good feature, and most of what is written about them online oversells what they are. A SharePoint agent is a scoped question-and-answer bot that sits on a site, a document library or a set of files you choose. It answers from that content, it respects the permissions on that content, and it took Microsoft roughly one click to make it appear. For a policy library, a project site or an onboarding hub, that is a real win with no procurement cycle attached.
One search · cited answers · permission-aware
The limit that decides whether it can be your company search is published in Microsoft's own documentation: you can include up to 20 source items as the knowledge source of an agent, and those source items can be sites, document libraries, folders and files. Twenty. That is a deliberate design decision, not an oversight, because a SharePoint agent was built to be narrow and precise rather than broad. The supported file types are equally specific: DOC, DOCX, PPT, PPTX, XLSX, PDF, TXT and RTF, with images, meeting recordings, videos and OneNote listed as coming later.
Then there is the licensing question, which is where most evaluations stall. To create an agent you need a Microsoft 365 Copilot license and the ability to add files to the site. To use one you need either that license or a pay-as-you-go billing policy your admin sets up as a resource in Azure. With neither, Microsoft's own comparison table says plainly that users cannot create or use agents. And because SharePoint agents are grounded in the tenant graph, Microsoft documents that each interaction with a SharePoint agent uses 12 messages, 10 for tenant graph grounding and 2 for the generative answer, billed through the Copilot Studio meter for anyone without a Copilot license.
InSearch is the alternative for the other requirement, the one that starts with someone asking a question and not knowing which of your dozen apps holds the answer. One search across Google Drive, Slack, Notion, Confluence, Gmail, Jira and Salesforce, a written answer with inline citations back to the exact source documents, item-level permissions enforced live at query time, and a published per-seat price with no meter and no Azure subscription in the setup path. If your knowledge fits in 20 SharePoint sources and everyone already has Copilot, use the SharePoint agent. If it does not, this is the gap it leaves.
SharePoint agents are billed pay-as-you-go, a model compared in SharePoint agents versus Copilot Studio agents. Because agents are scoped to sites, SharePoint search shows what a single query across Microsoft 365 and beyond looks like, and the Microsoft Copilot alternative covers the tenant-wide option.
SharePoint agents are scoped Q&A bots over up to 20 SharePoint sources, gated behind a Microsoft 365 Copilot license or Azure pay-as-you-go billing. InSearch is the alternative when the requirement is company-wide search across every app, cited and permission-aware, at a fixed per-seat price.
Side by side
SharePoint agents vs InSearch, honestly
A fair look at what each does well. Both are capable tools. Here is where they differ.
| What matters | InSearch | SharePoint agents |
|---|---|---|
| Sources it can read | Every connected app: Drive, Slack, Notion, Confluence, Gmail, Jira, Salesforce, with no cap on documents | Up to 20 source items per agent, and those items must be SharePoint sites, libraries, folders or files |
| Content outside Microsoft 365 | First-class. Slack threads and Notion pages are searched the same as Drive docs | Not covered. Agents answer from content in SharePoint sites, pages and selected files |
| File types | Documents, spreadsheets, slides, PDFs plus chat messages, email and ticket text | DOC, DOCX, PPT, PPTX, XLSX, PDF, TXT, RTF. Images, recordings, video and OneNote listed as coming soon |
| Who can create one | Nothing to create. Connect a source and ask | A user with a Microsoft 365 Copilot license and the ability to add files to the site |
| Who can use one | Anyone with a seat | Copilot-licensed users, or unlicensed users covered by a pay-as-you-go billing policy. With neither, no access |
| Cost shape | Published per-seat price. Seats times price is the bill | Included for Copilot-licensed users. Otherwise 12 messages per interaction on the Copilot Studio meter |
| Admin setup | Connect apps with OAuth. No cloud subscription required | For pay-as-you-go: an Azure subscription, a resource group, SharePoint admin plus Azure Owner or Contributor roles |
| Permissions | Item-level permissions inherited from each source and enforced live at query time | Honored properly. Content you cannot open is not used in the answer, even if it is in the agent sources |
| Best suited for | One search box for the whole company across a mixed tool stack | A precise assistant on one site or library, for a tenant already standardized on Copilot |
Comparison reflects general, publicly understood positioning. Capabilities change, so check each product for the latest.
Why teams pick InSearch
One cited, permission-aware search across every app
The cap is not the problem, the boundary is
Microsoft caps an agent at 20 source items and says you can nest data at a higher level if you need more, so the number itself is rarely the blocker. The real limit is that all 20 have to be SharePoint, in a company where the answer could be in any of forty places.
The bill depends on who is asking
Interactive use by Copilot-licensed users costs nothing extra. Everyone else runs 12 messages per question through the Copilot Studio meter, so the same agent has two completely different cost profiles depending on headcount licensing.
SharePoint is not where the answer usually is
The decision often lives in a Slack thread, the rationale in a Google Doc and the follow-through in a Jira ticket. An agent reading SharePoint alone returns a confident, partial answer, which is worse than no answer.
Good questions
SharePoint agents vs InSearch, answered
More comparisons
See how InSearch compares
GoSearch alternatives
Cited, permission-aware answers across every app.
vs DustDust alternative
Search-first, with cited answers out of the box.
vs GuruGuru alternative
Live search across every app, not just a curated wiki.
vs Amazon KendraAmazon Kendra alternative
Kendra is in maintenance mode. Here is where to land.
vs LucidworksLucidworks alternative
Workplace search without a Kubernetes cluster to run.
vs SinequaSinequa alternative
Built for the modern SaaS stack, not a two-year rollout.
vs Atlassian RovoAtlassian Rovo alternative
For companies that are not only an Atlassian shop.
vs HebbiaHebbia alternative
Company-wide cited search, not finance-desk document diligence.
vs OnyxOnyx alternative
The same cited, permission-aware answers, fully managed and no-ops.
vs Amazon Q BusinessAmazon Q Business alternative
Still open to new customers, and permission-aware by default.
vs Gemini EnterpriseGemini Enterprise alternative
Cross-app cited answers without a Google Cloud project to run.
vs Google AgentspaceGoogle Agentspace alternative
Agentspace became Gemini Enterprise. Here is what that means for buyers.
vs ServiceNow OttoServiceNow Otto alternative
Otto absorbed Now Assist and Moveworks. Here is the focused option.
vs Microsoft Copilot StudioCopilot Studio alternative
Build an agent, or just get the answer.
Find any answer at work with InSearch
One search across all your company apps, a clear cited answer, scoped to exactly what each person can see, and never trained on your data.
One search across every app · cited answers · permission-aware · never trains on your data