# How to Set Up the Confluence Connector for Microsoft 365 Copilot

> Set up the Confluence Cloud connector in the Microsoft 365 admin center so Confluence pages show up in Microsoft 365 search, Copilot and your Copilot agents.

- URL: https://www.danielanderson.io/articles/copilot-confluence-connector/
- Author: Daniel Anderson, Microsoft MVP (https://www.danielanderson.io/)
- Published: 2025-12-20
- Updated: 2026-10-19
- Video: https://www.youtube.com/watch?v=4oH1pQ6d16U

## Key takeaways

- The Confluence Cloud connector is set up in the Microsoft 365 admin center under Copilot, then Connectors, then Gallery.
- Once crawled, Confluence pages appear in Microsoft 365 search results with their own Confluence filter, and in Copilot answers.
- A Copilot agent can use the connector as a knowledge source, so an agent can answer from Confluence without anyone leaving Microsoft 365.
- By default the connector only shows Confluence content to people who already have access to it in Confluence.
- Permission changes in Confluence only reach Microsoft 365 in the daily full crawl, because the 15-minute incremental crawl skips them.

To connect Confluence to Microsoft 365 Copilot, go to the Microsoft 365 admin center, choose **Copilot**, then **Connectors**, pick **Confluence Cloud** from the **Gallery**, enter your Confluence URL and authenticate. After the first crawl, your Confluence pages show up in Microsoft 365 search and Copilot answers, and you can add Confluence as a knowledge source in any Copilot agent.

The connector keeps Confluence's own permissions. Space permissions and page restrictions carry across, so people only see Confluence content in Microsoft 365 that they can already open in Confluence. New and changed pages sync every 15 minutes by default, with a full crawl once a day.

## What do you need before you start?

- **Confluence Cloud.** This connector is for Confluence Cloud only. Confluence Server and Data Center use a separate on-premises connector.
- **Your instance URL.** It looks like `https://yourcompany.atlassian.net`. A Confluence admin can find it at admin.atlassian.com under **Product URLs**.
- **A Confluence admin.** You need one to create the OAuth app if you use your own, and to approve the connection. If your Microsoft 365 admin isn't also a Confluence admin, the authorisation step can fail with "Invalid connection credentials" until the Confluence admin approves it.
- **Matching identities.** By default, a person's Confluence email has to match their Microsoft Entra ID user name for permissions to line up.

## How do you add the Confluence connector in the admin center?

1. In the Microsoft 365 admin center, choose **Copilot**, then **Connectors**.
2. On the **Connectors** tab, choose **Gallery**, then **Confluence Cloud**.
3. Give it a display name. It defaults to Confluence Cloud, and it's what people see in Copilot references and as the search filter, so keep it recognisable.
4. Enter your instance URL.
5. Choose an authentication type: **OAuth 2.0**, which uses a Microsoft-managed app and is the recommended option, **Basic authentication** with a username and Atlassian API token, or **OAuth 2.0 with customized app** if you want to use your own Atlassian app.
6. Choose **Custom setup** if you want to change who sees results, what gets indexed or the sync schedule before it goes live. **Create** deploys it with the defaults.

When I recorded this I used my own OAuth app, which needs a client ID and secret from Atlassian. The Microsoft-managed OAuth 2.0 option now saves you that step, but if your organisation wants its own app, the next section covers it.

## How do you create the OAuth app in Atlassian?

I'm not a Confluence user, so I pasted a screenshot of the connector setup screen into Copilot and asked it to help me get a client ID and secret. It came back with a checklist, the developer console URL and the steps, which match Microsoft's documented process:

1. Sign in to the Atlassian developer console (developer.atlassian.com/console/myapps) with a Confluence admin account.
2. Choose **Create**, then **OAuth 2.0 integration**, name the app and create it.
3. Choose **Permissions**, then **Add** next to **Confluence API**.
4. Choose **Configure**, then **Edit scopes**, and add the read scopes the connector needs: content details, audit records, pages, groups, user details, spaces, content restrictions and space permissions, content summaries, comments, and attachments. **Save**.
5. Choose **Authorization** and add the callback URL `https://gcs.office.com/v1.0/admin/oauth/callback` (Government tenants use `https://gcsgcc.office.com/v1.0/admin/oauth/callback`). **Save**.
6. Choose **Settings**, copy the **Client ID** and **Secret**, and paste them into the connector.

Copy the scope codes from Microsoft's [Confluence Cloud connector deployment guide](https://learn.microsoft.com/en-us/microsoft-365/copilot/connectors/confluence-cloud-deployment) rather than typing them, because a missed scope is an easy way to end up with gaps.

## How do you choose who sees Confluence results and what gets indexed?

In **Custom setup** you've got three groups of settings.

**Users.** Access permissions default to **Only people with access to this data source**, which respects Confluence permissions. The other option, **Everyone**, shows all indexed content to everyone in your tenant. If Confluence emails don't match Entra ID user names, set up identity mapping here.

**Content.** All pages and blog posts are indexed by default. You can include or exclude spaces, or filter with a Confluence Query Language (CQL) string, then use **Preview results** to check it before saving.

**Sync.** Incremental crawls pick up new and changed content every 15 minutes, and a full crawl runs every day. You can change both. I left mine on the defaults.

## Roll it out to a small group before anyone else sees it

There's a **Rollout to limited audience** toggle in the deployment settings, and I'd use it every time. Pick a handful of people from a couple of teams, let them search and ask Copilot questions for a week, and check what comes back against what they'd expect to see in Confluence.

Most of the risk sits in permissions, so I'd leave access on **Only people with access to this data source** and treat **Everyone** as a deliberate decision someone signs off. A space that was opened up years ago and never tidied becomes far easier to stumble on once it's in Microsoft 365 search. Remember that permission changes only sync in the daily full crawl, so if you lock down a space in Confluence, it can take up to a day for that to reach Microsoft 365.

The other thing I'd do before a wide rollout is look at how the spaces are organised. Clear space names, labels and page descriptions help Copilot match the right page to a question, the same way metadata does in SharePoint. I make the case for that groundwork in [why your SharePoint needs an information architecture before AI](https://www.danielanderson.io/articles/sharepoint-information-architecture-before-ai/), and it applies to Confluence just as much.

## Where do Confluence results show up once it's connected?

**Microsoft 365 search.** A search for "how do I create a SharePoint page" brought Confluence pages back alongside SharePoint results, with a Confluence filter next to the SharePoint one. Hovering over a result and choosing **Overview** gives a summary of the page, with a link to open the original in Confluence.

**Copilot agents.** In a new agent's **Configure** tab, I added Confluence as a data source under **Knowledge**, and the agent answers from the Confluence content. If you're building agents with several jobs, [adding skills to Copilot agents in Agent Builder](https://www.danielanderson.io/articles/copilot-agent-builder-skills/) shows how to keep them in one agent. Other connectors, such as monday.com and Smartsheet, are set up from the same **Gallery**.

## What should you check after setup?

- **Index count.** On the connector's details page, check the index count roughly matches the number of pages in the spaces you included. A Confluence admin can pull a page count from the space reports.
- **Content that's left out.** User profiles, page history, archived pages, unpublished drafts, third-party app content and anything in the recycle bin aren't indexed.
- **Pages nobody can see.** Content without permission information is hidden from everyone, so a space with no space-level permissions won't show up for anyone.
- **Basic authentication tokens.** If the account behind the API token is deactivated or the token is revoked, syncing fails with a permission error until you enter a new token.

## Frequently asked questions

**Can I index only some Confluence spaces?**
Yes. Include or exclude spaces in the content settings, or use a CQL filter such as `lastModified >= "2024/12/31"` to sync only pages changed after a date. A CQL query that returns nothing gives you no error, just an empty preview, so check the preview before you save.

**Can I change which Confluence fields are searchable?**
Yes. In the content settings, each indexed property (Title, Labels, SpaceName, UpdatedOn and so on) can be set as searchable, queryable, retrievable or refinable, and you can change its semantic label and alias.

**Are Confluence attachments searchable in Copilot?**
Yes. Files attached to pages and blog posts are indexed into the page's content, which is why the attachment read scope is part of the app setup.

**Can I build agents in Copilot Studio with the Confluence connector?**
Yes. The connector can be a knowledge source for declarative agents in Copilot Studio, Agent Builder and the Microsoft 365 Agents Toolkit.

**Why can't some people see Confluence results they can open in Confluence?**
Usually identity mapping. If someone's Confluence email doesn't match their Entra ID user name, the connector can't match their permissions, so set up a mapping formula in the user settings and wait for the next full crawl.