What it does

Copilot connectors (formerly Graph connectors) index external content — ServiceNow, Confluence, file shares, databases, custom sources — into the Microsoft Graph so Copilot and Microsoft Search can ground answers on it.

Key facts

• Content is indexed and stored in your tenant's semantic index; it becomes citable by Copilot.
• Many prebuilt connectors from Microsoft and partners; you can build custom ones via the connector API.
• Indexed items count against a per-tenant item quota that varies by licensing — check the current limit before large loads.
• Access is trimmed to the source's permissions (ACLs), provided you map them correctly.
• Requires configuring crawl schedules and, usually, a service account.

When to use / skip

Use whenever Copilot needs to answer over content that lives outside Microsoft 365. Skip if the data is already in SharePoint/OneDrive, or if you only need one-off retrieval — an API plugin may fit better than a full index.

Configuration decisions

• Which sources, and how much of each to index.
• ACL mapping strategy — the part that goes wrong most often.
• Crawl frequency vs freshness needs.
• Schema and search/result settings.

Gotchas

• Get ACL mapping wrong and users either see content they shouldn't or nothing at all — test with real accounts.
• Item quotas are real; a large source can blow the tenant limit and stall other connectors.
• Reindexing large sources takes time; don't demo the morning after you kick off a crawl.

Consultant notes

• Connectors are the workhorse for grounding, but they're a licensing and cost conversation — item counts and index quota, not just effort.
• Governance: indexed external data now lives under Purview/DLP scope, which some data owners won't have expected.
• Budget properly for ACL design; it's where most connector projects lose time.

---

Review if item quotas or the connector gallery change materially.