How can I use rovo to query atlassian home projects and goals by tag?

jbarnes1

I'm trying to create an agent that will read an Atlassian Home topic page (filtered by a specific tag) to find all the projects and goals associated with that tag so that the agent can read through those projects/goals (and linked Jira/confluence) to find relevant status content.

I have tried multiple things with no success. The Rovo agent tells me that it doesn't have access to tag information through the public APIs. It also says that Atlassian Home is a single page app with no html payload, so it can't read content on Topic pages to identify linked work.

It's very surprising to me that the something as basic as tagging, which is absolutely supported in the graph, isn't accessible via API, even if it's not available via MCP. If needed, perhaps I could build a forge app, but I did experience similar blockers when I was trying to run python scripts from my CLI. 

If anyone can point me in the right direction, or definitively tell me there is no solution, that would be appreciated :-) 

Alexander Nilsson

Hello,

the Rovo agent gave you outdated information. Tags on Atlassian Home goals and projects are accessible through a public API, just not the one the agent is reaching for. The "Topic" UI in Atlassian Home is now called Tags, and it is fully queryable.

In January 2026, Atlassian released GraphQL APIs for Goals and Projects (Atlassian Home). The endpoint is https://{your-subdomain}.atlassian.net/gateway/api/graphql (you can also use <a href="https://team.atlassian.com/gateway/api/graphql" target="_blank" rel="nofollow noopener noreferrer">https://team.atlassian.com/gateway/api/graphql</a>), auth is Basic with a Base64 of email:api_token. The official announcement and getting started guide is here:

https://community.atlassian.com/forums/Goals-and-Projects-articles/New-GraphQL-APIs-are-available-for-Atlassian-Goals-and-Projects/ba-p/3171496

For your specific case, the goals_search query has a label field with aliases tag and topic, and the search field searches across goal name and tags. So a query like this returns every on-track goal with the tag "platform":

query {
  goals_search(
    containerId: "ari:cloud:townsquare::site/{siteId}"
    searchString: "label = \"platform\" AND status = on_track"
    first: 50
  ) {
    edges {
      node {
        id
        key
        name
        url
        owner { aaid }
        status { value }
        targetDate
      }
    }
    pageInfo { hasNextPage endCursor }
  }
}

The full field reference (including label/tag/topic, operators, and the 30-point cost budget) is here:

https://developer.atlassian.com/platform/goals/goals-graphql-api/search-queries/

The same pattern exists for projects via projects_search. Once you have the IDs back, goals_byId and projects_byId expose tags, contributors, linked work items, and linked goals/projects, so you can walk from a tag to the linked Jira issues and Confluence pages without scraping any UI. The byId schema is documented here:

https://developer.atlassian.com/platform/goals/goals-graphql-api/using-graphql-api/

Now, the awkward part. You asked specifically about doing this from a Rovo agent, and that is where it gets messy today. The Atlassian team confirmed in the announcement thread on January 22, 2026 that there is no MCP server implementation for the Goals and Projects GraphQL API yet, which is why Rovo itself cannot reach this data through its built in tools, and why your Rovo agent told you tags were unreachable. Forge access to the Goals and Projects APIs was also asked about on January 27, 2026 and there is no public timeline from Atlassian yet.

So the practical path right now is to call the GraphQL API from outside the Rovo agent context (a small Python or Node script, an n8n flow, or a Forge function using api.fetch with stored credentials) and either consume the result in your downstream system directly or feed it back into Rovo as context. Avoid trying to make Rovo "read the Atlassian Home page" itself, you are right that it is a SPA and that approach will not work.

Greetings,

Alex