Inspiration

Every sprint, the same three questions eat the standup: what went quiet, what's blocked, and who's drowning? Scrum masters dig through the board by hand to answer them — and by the time the answers are assembled, they're stale. We wanted those answers in one chat message, and the follow-up done automatically.

What it does

Sprint Scout is a Forge app with three faces:

  1. A Rovo agent you chat with in Jira. Ask "How healthy is our current sprint?" and it answers with progress, days left, and an on-track / watch / at-risk signal — with the reason, not just a number. Ask what's blocked or stale and it lists the items with assignees, days idle, and blocking keys. Ask who's overloaded and it reads team workload from the Teamwork Graph. And it does work, not just reports: "Draft our sprint review notes" creates a Confluence page (shipped / open / stale / blocked / hotspots / discussion prompts), and "Nudge the stale ones" posts a polite status-request comment — with a real @mention — on each stale item. Both write actions ask for confirmation first.
  2. A Jira dashboard (global page) with the same health picture as a live view, saved settings, and one-click review drafting.
  3. A Jira issue panel giving a one-glance verdict on the work item on screen — healthy, stale, blocked or done — with a one-click nudge. The four read actions are also exposed as MCP tools (rovo:mcp), so any agent in Rovo Studio — or any MCP-compatible client an admin enables — can call them.

How we built it

One Forge app, all Atlassian-native — no external services, no paid APIs.

  • Sprint data: Jira Agile API (boards → active sprint → sprint issues), with a JQL fallback to all open work when there's no active sprint.
  • Team workload: Teamwork Graph curated queries (atlassianUserInTeams → atlassianTeamMembers → atlassianUserAssignedJiraWorkItems), paged 25 rows at a time with retries, operation names following the [company]_[operation] rule.
  • Agent: rovo:agent with 6 actions (4 GET, 2 CREATE) and a prompt that forbids guessing, requires confirmation before writes, and reports partial reads as "may be incomplete".
  • UI: UI Kit (@forge/react) global page + issue panel, backed by one shared service layer — so the agent and the dashboard can never disagree.
  • Safety: everything runs as the user, so every answer is permission-filtered to what that person can already see.

Challenges we ran into

  • Rovo's practical limits: actions must answer well inside ~20 s and ~1 MB, so workload caps at 12 members × 1 page each and sprint issues cap at 200.
  • Teamwork Graph quirks: errors arrive as HTTP 200 with an errors array; a rate-limited call can return zero rows; a missing scope drops rows silently. The client treats all three as "could not read", never as "nothing there".
  • Status names vary by site, so "done" is detected by status category first, names second.
  • Note: the stale threshold is one saved setting ("Stale after (days)") that Rovo, the review draft, and the dashboard all honor. The demo video was recorded before this fix, when Rovo and the draft still used the 3-day default. ### Accomplishments that we're proud of
  • A complete product, not a demo: agent + dashboard + panel + MCP, sharing one analysis pipeline, with 19 unit tests.
  • The agent acts: review drafts and nudges close the loop instead of just reporting.
  • Honest answers: partial data is labelled partial, never presented as fact.

What we learned

  • The Teamwork Graph's curated-query model: typed root fields, per-object scopes, and paging discipline.
  • Designing for an agent consumer is different from designing for a human: action descriptions are the UI, and the prompt is the product spec.
  • rovo:mcp turns one app's actions into tools every other agent can reuse.

What's next for Sprint Scout

  • A scheduled Monday-morning digest (sprint health posted to the team's Confluence space or as a Jira comment).
  • Sprint-over-sprint trends (velocity, recurring stale items) stored in Forge storage.
  • A "suggest next action" step: propose reassigning overloaded members' work, for the user to approve.

Setup steps for judges

  1. Open the installation link above and install Sprint Scout on a Jira Cloud site (an org admin must approve the Teamwork Graph scopes).
  2. You need a project with an active sprint (or any project with open work — the app falls back to all open work and says so). For the workload question, be a member of an Atlassian team with assigned work.
  3. Rovo agent: open any work item → Open Rovo Chat → Change agent → Browse agents → "Sprint Scout". Try the conversation starters, e.g. "How healthy is our current sprint?" (Rovo requires a paid plan or a Standard/Premium trial.)
  4. Dashboard: Apps → Sprint Scout. Save your project key (and a Confluence space key for review drafts), then Refresh health.
  5. Issue panel: open any issue → ••• menu → Sprint Scout.
  6. MCP tools: in Rovo Studio, add the "Sprint Scout MCP" tools to a custom agent. External MCP clients: an admin enables exposure in Atlassian Administration → Connected apps; endpoint https://mcp.atlassian.com/forge/.
  7. To see a review draft: ask the agent "Draft our sprint review notes in the [SPACE] space" (or use the dashboard button) and confirm.

Installation link

Install Sprint Scout on your Jira Cloud site: https://developer.atlassian.com/console/install/a7e3c905-c264-4d1d-a1f2-8b97a832227c?signature=AYABeJQhiEBvOx8qWY3HYIHmUTUAAAADAAdhd3Mta21zAEthcm46YXdzOmttczp1cy13ZXN0LTI6NzA5NTg3ODM1MjQzOmtleS83MDVlZDY3MC1mNTdjLTQxYjUtOWY5Yi1lM2YyZGNjMTQ2ZTcAuAECAQB4CxqLTo9l4lqXuFjNim1xJM%2Bp%2F3TVP9xHS2BrU5DRRucB6tzAForNLe9lcooabxR%2BtAAAAH4wfAYJKoZIhvcNAQcGoG8wbQIBADBoBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDHePYWnqecaTrAe7eAIBEIA7jwLMe8OgeMxB8g6KaPlMxARgE22KklNqC2nUPF6ZbMp4f9cxXcidaQxUdQ9k1Zo01f%2FIEub%2FXzNNRiEAB2F3cy1rbXMAS2Fybjphd3M6a21zOmV1LXdlc3QtMTo3MDk1ODc4MzUyNDM6a2V5LzQ2MzBjZTZiLTAwYzMtNGRlMi04NzdiLTYyN2UyMDYwZTVjYwC4AQICAHi9bVMAkme4qVQX1fQHo6ePw3H40A7cGfz7yrmSfBWMXgHqwrMEqUaCvU2dmO4oTTEMAAAAfjB8BgkqhkiG9w0BBwagbzBtAgEAMGgGCSqGSIb3DQEHATAeBglghkgBZQMEAS4wEQQMxkvwWPOvToiaZ3mrAgEQgDuI%2BLdjI6donL19Cb0r9YlOF9p94Qe%2BEm1eaAktZpklt4o84T1WAVwmBkbYGEoBBTmx9SzXWGNQppUJOAAHYXdzLWttcwBLYXJuOmF3czprbXM6dXMtZWFzdC0xOjcwOTU4NzgzNTI0MzprZXkvNmMxMjBiYTAtNGNkNS00OTg1LWI4MmUtNDBhMDQ5NTJjYzU3ALgBAgIAeDYASx0XCjkny4iLV6vz%2BKtEGSGT6kdt0QRpkn6D%2FlVEAfAdjX%2BQWK9nsghDu2IgK%2FsAAAB%2BMHwGCSqGSIb3DQEHBqBvMG0CAQAwaAYJKoZIhvcNAQcBMB4GCWCGSAFlAwQBLjARBAzZ%2FnXJ4H8SYJw8nTwCARCAOxXK75hp8U8EHUjQddL9IiVBknq9sHJSIb5iGOit9%2FFek3umycjwwTkBGiZdzVvBLzDZt%2FZHc1DpNILuAgAAAAAMAAAQAAAAAAAAAAAAAAAAAPm6vNSoF644RbVvIHK8u%2Bj%2F%2F%2F%2F%2FAAAAAQAAAAAAAAAAAAAAAQAAADI4fFSgS9Ofx9KYnco4UIsQPh2sZkCI9Kq7TQhQS3q3fTEO9MHrVo6H%2F%2F3%2BVarmBJX%2F72M3wsU%2FUIltBxP6Zd0Noso%3D&product=jira

Built With

Share this project:

Updates

Submission history