Combining Claude, Atlassian MCP, and Rovo

[#1]

I posted a few months ago about some of the Rovo agents I created, and what I thought of the experience.

Like everyone else on LinkedIn, I’ve been deep in Claude exploration, and this time I want to share how I’ve combined Claude and Rovo or the Atlassian MCP for two processes.

Finding what’s waiting for me

I get pinged on a lot of Confluence pages and Jira issues. “@Laura what do you think?” “@Laura is this okay?” “@Laura just updated”. Three days later, I remember someone is waiting for my feedback, just not where and for which topic! I don’t love the notification feed as it tends to get cluttered with non-essential updates (“Your Loom was watched!”), and I’m drowning in notification emails.

A long time ago I tried setting up a filter in Jira to catch all the times someone mentioned me in a comment, but this only works if my name is in text, not a user mention, because @mentions are stored as structured "mention" nodes in Atlassian Document Format, not in searchable text. There's no JQL field like comment-mentions = currentUser() or mentioned-in-comment = accountId.

How Claude did the heavy lifting querying IDs & applying regex

Claude uses the MCP Atlassian connector to run a JQL query searching for open issues where I’m mentioned, and then checks the comments of each issue. It can extract plain text recursively from content nodes, treating type: "text" nodes as text and type: "mention" nodes as @displayName.

Then it runs a CQL query with mention = "accountId" (since Confluence supports this), and checks if the comment is resolved or not.

Claude then applies its own regex-based filtering to figure out which mentions are actually asking me to do something vs. just name-dropping me.

Then it age-buckets the results, displays it in a nice html format with links to the page or Jira work item 😀 I didn’t know where to start to build something on my own that could analyze Atlassian Document Format or apply regex-based filtering; thankfully Claude built it for me!

Building content briefs that include curated reference pages from Confluence

Preparing a content brief for someone to write a blog or article requires completing keyword research, Jira/Confluence best practice research, and then also research about the product I want to include in the article.

Claude can help with the first 2 types of research, notably with the MCP server from our keyword tool, but it tends to have a hard time finding the right information about my company's products on it’s own. Searching the internet leads to too many hallucinations or just incorrect conclusions.

But I already built a Rovo agent to answer questions about our products! This Rovo agent has custom knowledge with curated pages, so I know the results it gives will be much more accurate than if I just searched our entire Confluence site.

So I just included a query to our Rovo agent as part of the Claude skill to get the right information. Right now there’s no API to query Rovo agents, so I have to use Claude in Chrome, but it works pretty well. Claude opens a new tab, enters it’s question, and takes the answer from Rovo for the next part of the process.

Once Claude has fetched the information from Rovo, it combines all the research into a nice format, and asks me where I want to create a Confluence page.

Currently I find it is seamless to create new pages under parent pages, but Claude sometimes struggles to create pages under folders and has to do several attempts.

Circle back in 6 months?

Things have changed dramatically since I wrote my other article about creating Rovo agents. I’m sure in 6 months time the way Claude and Rovo or the Atlassian MCP can be used together will allow for vastly more sophisticated processes.

[Colton_Kosicek]

Hey @Laura Campbell  - love to see what you are doing.  What strategic value are you seeing by leaving the Rovo agents as part of the lifecycle instead of curating the same within a specific Claude skills?  It feels like Rovo will always know "atlassian" better then claude but it leaves the door open to double the updates (keep Rovo agent updated and tested and test the Claude skills). Is the claude skills (Pardon my ignorance not integrated claude and atlassian yet to know) not capable of re-using some of the Rovo logic on it's own?

Is there a way of creating a simple 1 pager on confluence to describe your products to solve that problem or is this a "no alt solution" wall that you are mindfully hurdling with skill + Agent.

[John Johnson]

This is a really insightful breakdown of how Claude and Rovo can work together, especially for managing notifications and building structured workflows. The way you handled mentions and filtering with regex is smart, and it shows how much time automation can save in daily tasks.

Also, balancing productivity with downtime is important too. After dealing with complex workflows like this, some people prefer simple ways to relax, like using platforms where they can Watch Anime Episodes on Mobile and unwind after work.