Inspiration

Every Jira team has asked the same question: "Who should work on this?" The answer usually comes from memory, a guess, or whoever happens to be online. The person who wrote the docs, the one who fixed the same bug last quarter, or the component owner is often someone nobody thinks to ask — and the person who does get asked is often already overloaded. I wanted a way to answer that question with evidence instead of guesses.

What it does

KnowWho is a Rovo agent that ranks the right reviewer, owner, or unblocker for any Jira issue, with a transparent score breakdown. Ask it "Who should review this SALES-1?" and it traverses the Atlassian Teamwork Graph, reads Confluence authorship, finds who resolved similar issues before, identifies the component owner, checks each candidate's current workload, and returns a ranked list. Every point is explained and linked to its source — no black box.

The same five capabilities are also exposed as MCP tools, so any MCP client can ask the same questions. A Jira issue panel shows the ranking in context, with a button to assign the top expert.

How I built it

KnowWho is a Forge app (Node.js 22). The Rovo agent exposes five read-only actions — rank experts, find page authors, find past resolvers, find the component owner, and check workload — backed by the Teamwork Graph with a REST fallback. I kept the actions read-only on purpose, so the agent stays assignable from Jira Automation rules. The ranking uses a published, weighted formula: Confluence authorship +3, past fixes +2, component lead +4, recency +1, same team +1, and an overload penalty of −1 per open item over three. Role tweaks the weights: a reviewer should know the area, an owner should own it, an unblocker should be active and available.

Challenges I ran into

The Teamwork Graph is in Preview, so I could not verify every field shape in advance — I built a REST fallback for each signal so the agent still works when the graph is unavailable. The Forge manifest lint caught several issues (a description over 255 characters, a missing panel icon, a YAML colon that broke parsing), which I fixed and redeployed. The Jira issue panel did not render on the demo site's business-project issues, so the demo video focuses on the Rovo agent, which works end to end — 17 steps to a ranked, evidence-based recommendation.

What I learned

I learned how the Teamwork Graph connects people, work, and knowledge across Jira and Confluence — and that a transparent ranking earns more trust than a clever one. If the team can see exactly why someone ranks first, they can disagree with reasons instead of vibes. I also learned that a read-only action set is a feature, not a limitation: it keeps the agent usable from Automation and safe to deploy broadly.

Try it

Source code: https://github.com/yovmiar19/knowwho Demo video: https://youtu.be/QGJS9n5LMgs

Built With

  • atlassian
  • forge-(node.js-22)
  • jira-and-confluence-rest-apis
  • rovo-agent-+-mcp
  • teamwork-graph-api
  • ui
Share this project:

Updates

Submission history