Govern external AI access with the Rovo MCP server
10 min
By the end of this lesson, you'll be able to:
- Explain what the Rovo MCP server is
- Describe how external AI tools connect to the Rovo MCP server
- Explore how you authorize an AI tool to access your app data
- Manage how third-party AI agents and the Teamwork Graph CLI access your organization's data
What is the Rovo MCP server?
The Rovo MCP (Model Context Protocol) server is a secure gateway for Atlassian Cloud. It lets approved external tools, such as AI assistants, search, summarize, and update data in Jira, Confluence, and Compass. These tools use the same permissions and security rules you already have in Atlassian Cloud.
👇 Click the boxes below to explore how different tools use the Rovo MCP server.
External AI tools use your current permissions to take actions in Jira, Confluence, and Compass. You should only give users the access they need to do their work.
How data flows through the Rovo MCP server
Learning how data flows through the Rovo MCP Server helps you confidently adopt external AI tools without exposing sensitive Jira, Confluence, or Compass data.
Here is how the data flows:
- An external AI tool sends a request to the Rovo MCP server.
- The server checks if the tool is allowed, confirms your organization allows the action (read, write, or search), and confirms the user's permissions.
- The server gets or updates the data in Jira, Confluence or Compass and sends the result back to the AI tool.
The server only lets you see data that you already have permission to view in Atlassian Cloud. It follows the same security rules and roles you already have in your spaces.
Authorize external AI tools to access your apps through the Rovo MCP server
Control how tools connect to your site
As an administrator, you decide how external AI tools access your organization's data. Authorizing the right tools helps your teams work more efficiently. It also makes sure that only trusted tools can access your Atlassian data.
👇 Click the boxes below to explore how you can control which tools connect to your Atlassian apps.
Authorize a domain
Organization admins control which AI tools can connect to the Rovo MCP server in Atlassian Administration by authorizing domains. Domains are the web addresses AI tools use to communicate with Atlassian. When an admin authorizes a domain or changes its settings, that change applies to all users in the organization.
You can find a list of Atlassian-supported domains that are already optimized to work with the Rovo MCP server in the Add domain menu. Explore authentication options at Available Atlassian Rovo MCP Server domains.
To authorize a domain and allow a new AI tool to connect to Rovo MCP server:
- Go to Atlassian Administration and select your organization.
- Expand Rovo.
- Select Rovo MCP server.
- Select Add domain and enter the address (for example: claude.ai).
- Select Add.
👉 For example: Marta is an admin who wants to make sure that users can only connect Cursor to the Rovo MCP server to access Atlassian data. In the Rovo MCP server settings, she turns off Allow Atlassian supported domains. She then goes to the Your domains section and adds only the Cursor domain. Now, AI tool calls from Cursor are allowed, while calls from other external AI tools are blocked.
👇 This is where the Rovo MCP server settings are controlled.
Once you add the domain, users can connect their external AI tool to the Rovo MCP server. To do this, they enter the Rovo MCP server URL into the tool's settings then sign in to their Atlassian account when prompted.
Manage permissions
You can control what external AI tools are allowed to do with your Atlassian data using the Permissions tab. This lets you allow or block access by intent Read, Write, or Search across specific apps like Jira, Confluence, and Bitbucket.
The Permissions tab is the authoritative control for MCP server access. It overrides any previous settings in Connected Apps or individual Marketplace apps.
To configure permissions:
- Go to Atlassian Administration and select your organization.
- Expand Rovo.
- Select Rovo MCP server.
- Select the Permissions tab.
- Allow or block access by intent for each app.
Manage authentication
👇 Click the tabs below to explore the two ways external tools can authenticate to the Rovo MCP.
OAuth 2.1 (Default) is the default sign-in method. It uses an Atlassian consent screen, so users never share their Atlassian passwords. Tools receive access tokens that expire and can be revoked at any time.
By default, users connect their AI tools using OAuth 2.1. However, you can allow them to use API tokens.
If you allow users to connect using API tokens, they can connect from any domain, even if you have not added it to your domains list in the Rovo MCP server settings. However, their connection must still come from a trusted IP address on your organization's allowlist.
To enable API tokens authentication for AI tools connecting to the Rovo MCP server:
- Go to Atlassian Administration and select your organization.
- Expand Rovo.
- Select Rovo MCP server.
- Toggle API token.
How IP allowlists affect the Rovo MCP server
If your organization uses IP allowlists, it adds an extra layer of security to your Rovo MCP server. An IP allowlist is like a guest list for your network. It only lets people access your data if they are connecting from a trusted location, like your office or a VPN. For an AI tool to connect, it must come from an authorized domain and a trusted IP address that complies with your IP allowlist configuration.
An AI tool call to the Rovo MCP server will only succeed if three conditions are met:
✔ The authentication method must be enabled in your Rovo MCP server settings (for example, you have enabled API tokens).
✔ The request comes from a domain you’ve explicitly allowed in the Atlassian Rovo MCP server settings.
✔ The request originates from an IP address that’s permitted by your organization’s IP allowlist (if IP allowlists are set up).
✔ The action (read, write, or search) is allowed for that app on the Permissions tab.
Allow third-party AI agents to call Rovo
Agent2Agent (A2A) lets third-party AI agents, such as Google Gemini, call Rovo to complete tasks using your Atlassian data. This is different from the Rovo MCP server: with MCP, external tools access your apps directly; with A2A, external agents call Rovo.
A2A is off by default. When an admin turns it on, the change applies to the whole organization. Third-party agents can only use Rovo in apps where Rovo is turned on. Users must also sign in with their own Atlassian account, so agents can only access data the user already has permission to view.
To allow third-party agents to call Rovo:
- Go to Atlassian Administration and select your organization.
- Expand Rovo, then select Agent2Agent.
- Turn on Allow A2A.
Control Teamwork Graph CLI access
The Teamwork Graph CLI lets developers give their AI coding agents access to Atlassian data from the terminal. As an admin, you choose which permissions the CLI can use in your organization. The Teamwork Graph CLI doesn't have an on/off toggle. To block access, clear all permissions and select Save and revoke sessions. This signs out anyone who's currently using the CLI. All CLI activity is recorded in your organization's audit log.
To manage Teamwork Graph CLI access:
- Go to Atlassian Administration and select your organization.
- Expand Rovo, then select Teamwork Graph CLI.
- Select the permissions the CLI can use, or select Allow all permissions by default.
- Select Save and revoke sessions to sign out existing sessions.