Announcing GA of the Atlassian Rovo MCP server 🎉

[#1]

We’re excited to share that the Atlassian Rovo MCP server is now generally available.

If your teams already work in AI assistants like Claude or ChatGPT, the Atlassian Rovo MCP server gives you a secure, admin-controlled way to let those AI clients take action in Atlassian products including Jira, Confluence and Compass (with Jira Service Management and Bitbucket Cloud on the roadmap).

Instead of copy/pasting between an AI chat and Atlassian, you can turn conversations directly into specs, plans and updates - while Atlassian remains the system of record.

What is the Atlassian Rovo MCP server?

The Atlassian Rovo MCP server is our Remote Model Context Protocol (MCP) server that exposes Jira, Confluence and Compass as structured tools to third-party AI clients.

From within an external AI client’s connector gallery, users can:

• Connect Atlassian Rovo MCP server to their AI client

• Consent to scopes and sign in with their Atlassian account

• Use tools backed by the Atlassian Rovo MCP server, including:

  • Rovo search and fetch to power deep research

  • Create and update Confluence pages

  • Create and link Jira epics and issues

  • Work with Compass entities

All actions respect users' existing permissions, so AI clients can only see and change what the user could.

👉 Product overview & docs:
https://support.atlassian.com/atlassian-rovo-mcp-server/docs/use-atlassian-rovo-mcp-server/

What can you do with it?

Here are a few common “chat-to-work” flows we see:

1. Turn a chat into a Confluence spec

   • Brainstorm and refine a product spec in Claude or ChatGPT

   • Ask it to create a Confluence page in the right space

   • The MCP server will:

     • Find the right space using Rovo search

     • Create a structured page (title, sections, etc.)

     • Share the page link back in the chat

2. Create Jira epics and issues from that spec

   • From the same conversation, ask:
     “Create an epic and related stories for this spec and link them to the page.”

   • The MCP server will:

     • Find the right Jira project

     • Create an epic and child issues

     • Link them back to the Confluence page

3. Understand a service using Compass

• In your AI client, paste a link to a service or describe it (e.g. “our checkout service”)

• Ask: “What is this service, who owns it, and how healthy is it?”

• The MCP server will:

  • Look up the Compass component and its owners

  • Pull key health, dependency, and runtime details

  • Summarize it back in the chat with links to Compass for deeper investigation

What’s new with GA?

Our GA focuses on making the Atlassian Rovo MCP server enterprise-ready:

• ✅ Audit logs for tool invocations

• ✅ Domain allowlist so admins can choose:

  • Allow any supported AI domain

  • Use Atlassian’s trusted list (with the ability to extend it)

  • Block all domains unless explicitly allowlisted

• ✅ Respects existing IP allowlists

• ✅ CSS support and runbooks in place

• ✅ Security whitepaper in the Atlassian Trust Portal

• ✅ Apps that deliver tailored UI experiences on top of MCP tools

• ✅ Atlassian Skills in Claude's directory

We’d love your feedback

If you’re already experimenting with AI clients and our MCP tools:

• What workflows are you trying to automate first?

• Which AI clients and Atlassian products are most important to you?

• What additional tools or controls would make the Atlassian Rovo MCP server more useful?

Please share your questions, feedback, or examples in the comments below so we can continue to evolve the Atlassian Rovo MCP server with you.

[Yanqing Lin]

excited to see it is in GA!

[Josh]

Thank you for the helpful post, @Jemma Swaak .

You mentioned GA will include "Audit logs for tool invocations" - can you please clarify if this includes the following?

1. Displaying unsuccessful attempts in addition to successful attempts
2. Including the source domain for each attempt (successful or unsuccessful)
   1. It was basically impossible for us to troubleshoot trusted domain configurations without this information.

Will the GA release include the feature enabled or disabled by default?

• The beta was launched with MCP enabled by default, which was a major pain point for us. We understand the desire to drive user adoption of new features, but that needs to be balanced against security considerations.

Our team had some conversations (MCP-871) with Jeffin and Erick about these topics back in November.

[aerbe_micron_com]

Thank you. 

You say Bitbucket is on the roadmap.  What's the timeline for this.  
[BCLOUD-23748] MCP server for Bitbucket cloud - Create and track feature requests for Atlassian products.