The Atlassian Rovo Model Context Protocol (MCP) Server is now GA for Jira, Confluence, and Compass. It lets AI tools and IDEs securely read and write Atlassian Cloud data using the MCP, with enterprise‑grade controls, including domain allowlists, IP allowlist support, and audit logs.
To start using Atlassian Rovo MCP Server:
Configure MCP settings in Atlassian Administration.
This includes Atlassian-supported domains, trusted domains added by admins, and security policies. For more details, see Control Atlassian Rovo MCP server settings and Available Atlassian Rovo MCP server domains.Check your network setup, if you use IP allowlisting.
Ensure the egress IPs used by your AI tools are included in your Atlassian Cloud IP allowlists so MCP tool calls aren’t blocked. For more details, see IP addresses and domains for Atlassian cloud apps.Connect an MCP‑compatible client.
From your chosen AI tool (for example, ChatGPT or Claude), configure a connection to https://mcp.atlassian.com/v1/mcp and complete the browser‑based OAuth 2.1 flow with your Atlassian account.Start using your tools.
Once connected, you can use natural‑language prompts to search, summarize, and update Jira issues, Confluence pages, and Compass components using supported tools. See Use Atlassian Rovo MCP Server | Atlassian Rovo MCP Server Cloud for more details.
This release also includes the following fixes:
Resolved intermittent “invalid token” errors
We’ve fixed an issue where some Atlassian Rovo MCP Server sessions could fail with “invalid token” or similar authorization errors, causing tool calls to stop working until users re‑connected. Token handling has been hardened so sessions are more reliable, especially for long‑running workflows.Improved reliability in authentication flows
Alongside the invalid token fix, we’ve made smaller stability and error‑handling improvements across our OAuth flows so that MCP connections recover more gracefully from transient failures.
Key changes
General availability: Atlassian Rovo MCP Server has graduated from beta and is now GA for Jira, Confluence, and Compass across Atlassian Cloud.
Domain allowlists for AI clients: In Atlassian Administration, admins can now explicitly control which external AI tools and domains are allowed to connect to MCP. Domains include both Atlassian-supported domains and any domains that you authorize access to MCP.
IP allowlist‑aware tool calls: MCP now respects your Atlassian Cloud IP allowlists. Users may be able to complete OAuth from any network, but actual tool calls are evaluated against your organization’s configured IP ranges.
Audit logging for MCP activity: Key MCP events, such as tool invocations and first‑time app installs, are now logged so admins can monitor usage and investigate issues.
Streamable HTTP /mcp endpoint: A new streamable HTTP endpoint (/v1/mcp) improves compatibility with modern AI clients while /sse remains supported for backwards compatibility. We recommend updating any custom clients configured to use /sse so they now point to /mcp.
Expanded tool coverage for Jira, Confluence, and Compass: GA includes a consolidated tool set for reading and writing Jira issues, Confluence pages, and Compass components (searching, creating, updating, linking, and more).
Reliability and token‑handling improvements (fixed): We’ve addressed issues related to invalid tokens that could interrupt MCP sessions, improving stability for long‑running workflows.
Updated docs, security guidance, and support model: MCP documentation has been updated to remove beta labelling, explain security and network behavior, and clarify how to get help via Atlassian Support and the Community.
Benefits
Bring Atlassian into your AI workflows: Let AI tools and IDEs securely read and write Jira, Confluence, and Compass data without copying content between systems.
Enterprise‑ready controls: Combine domain allowlists, IP allowlists, and audit logs so your security and compliance teams can govern which tools are allowed and how they’re used.
Consistent permissions model: MCP uses OAuth 2.1 and always respects existing Jira, Confluence, and Compass permissions—users only access what they can already see in Atlassian Cloud.
Better performance and compatibility: The new streamable HTTP endpoint /mcp works with a broad range of MCP‑compatible clients while preserving support for existing /sse configurations.
Operational readiness: A dedicated support model, updated troubleshooting guides, and clearer security guidance make it easier for admins to roll out MCP to teams at scale.
Reason for the change
Many enterprise customers want to safely use AI tools and IDEs with their Atlassian data while maintaining strict security, compliance, and audit requirements. Moving Atlassian Rovo MCP Server to GA—and adding controls like domain allowlists, IP‑aware enforcement, and audit logging—enables organizations to adopt MCP with confidence and bring Atlassian into their existing AI workflows.
How to prepare for the change
Review security and access policies: Work with your security team to define which AI tools and domains should be allowed to connect to MCP, and configure these in Atlassian Administration: https://support.atlassian.com/security-and-access-policies/docs/control-atlassian-rovo-mcp-server-settings/
Validate IP allowlists: If you use IP allowlisting for Jira, Confluence, Compass, or Rovo, confirm that the egress IPs for your AI tools and proxies are included so MCP calls are not blocked.
Align internal guidance: Share internal guidance with users on permitted AI tools, what data they can access through MCP, and expectations for responsible use.
Update custom or self‑hosted clients: If you’ve built custom clients that connect to MCP via the /sse endpoint, plan to update them to use the new /v1/mcp endpoint for better compatibility going forward.
Set up monitoring and support: Ensure the right admins have access to MCP audit logs and that your teams know how to raise support requests (specifying “MCP”) and share feedback via the Atlassian Community:
Recommended Comments
Create an account or sign in to comment