The problem

When a teammate goes on leave, their work stalls because the context lives in their head and in scattered tickets. Nobody knows what they owned, who to ask, or what was already decided. Handover builds one pack before they leave, and answers questions about it while they are gone.

How it works

Pick the person and their first day away. Handover reads their open Jira work, their Confluence pages and the Teamwork Graph around them: their teams, the people involved, the decisions already made. High priority work and anything due in the first week goes to urgent. The rest can wait. The pack is a generated summary, not the person's own words, and a human checks it before anyone relies on it.

Three parts

(1) A Jira project page that builds the pack, with Urgent, Can wait, Who to ask, Decided already and Docs to read. (2) A Rovo agent, Handover Guide, that answers questions about the pack while the person is away, citing the pack and the original tickets, and saying plainly when a source could not be read instead of saying there is nothing. (3) An MCP server, hosted by Atlassian, exposing the same three actions (fetch-handover-pack, fetch-person-work, list-handover-contacts) to any MCP client, with the caller's own permissions on the site.

Built with

Forge, Teamwork Graph API, Rovo agent, Forge MCP, Jira, Confluence.

How to test

Install the app from the installation link below (Jira and Confluence). It needs a site admin to install. Open any Jira project and use the Handover page: pick a teammate and a leave date, then click Build handover pack. Ask the Handover Guide agent in Rovo chat about the pack. The Rovo agent needs a Standard or Premium trial site; the pack page and the graph queries work on a free site. The demo video shows the pack and the agent working on a seeded demo site.

Judges install the app from this link: https://developer.atlassian.com/console/install/1c82e9f4-2072-4818-9ac9-1c978c2718fd?signature=AYABeKTeheWd7von8VK8GlHBqpcAAAADAAdhd3Mta21zAEthcm46YXdzOmttczp1cy13ZXN0LTI6NzA5NTg3ODM1MjQzOmtleS83MDVlZDY3MC1mNTdjLTQxYjUtOWY5Yi1lM2YyZGNjMTQ2ZTcAuAECAQB4CxqLTo9l4lqXuFjNim1xJM%2Bp%2F3TVP9xHS2BrU5DRRucBWnKIE3IqRKERVODMVPun8wAAAH4wfAYJKoZIhvcNAQcGoG8wbQIBADBoBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDCC4gNVzuNtPjSnWWAIBEIA76jmh24orZAS3oPjRsJZJZuxULNDBaWUceedBuUErfUGxT45cWm3raQfuigQ1WU14DHpu2wUEDOi%2FJl8AB2F3cy1rbXMAS2Fybjphd3M6a21zOmV1LXdlc3QtMTo3MDk1ODc4MzUyNDM6a2V5LzQ2MzBjZTZiLTAwYzMtNGRlMi04NzdiLTYyN2UyMDYwZTVjYwC4AQICAHi9bVMAkme4qVQX1fQHo6ePw3H40A7cGfz7yrmSfBWMXgHgdDqYoL9r0jEg3PwOPmMHAAAAfjB8BgkqhkiG9w0BBwagbzBtAgEAMGgGCSqGSIb3DQEHATAeBglghkgBZQMEAS4wEQQM5ZcjZ4ENd8PHPLfKAgEQgDtl2JVsumzUKm2CM3DhOS5ol9Ef7uXgQVGaRlcy3u2TxmXiiYVc0HZ66I1bpdWNYAt6%2F64uiSMFSjc63QAHYXdzLWttcwBLYXJuOmF3czprbXM6dXMtZWFzdC0xOjcwOTU4NzgzNTI0MzprZXkvNmMxMjBiYTAtNGNkNS00OTg1LWI4MmUtNDBhMDQ5NTJjYzU3ALgBAgIAeDYASx0XCjkny4iLV6vz%2BKtEGSGT6kdt0QRpkn6D%2FlVEAaon7F4qdJIWJo9MfTpVjnUAAAB%2BMHwGCSqGSIb3DQEHBqBvMG0CAQAwaAYJKoZIhvcNAQcBMB4GCWCGSAFlAwQBLjARBAzBRYGJV33B%2FeMd1zECARCAO%2BeBMcLUlKKSHa52pcBRjHxa52TjbG1ydtm9s4Rvu930jBCT2KSRCOE%2FbIYFRS8EZFLkDYKbkRrYKcZKAgAAAAAMAAAQAAAAAAAAAAAAAAAAACV055nJqG6nuhL0dswfsaz%2F%2F%2F%2F%2FAAAAAQAAAAAAAAAAAAAAAQAAADKZDHQOLhWB%2FTlvqvIdCKDWTRbm91gDIDyCKC7l3rw7kW6PrPCvXj5bmdt1hzfpjpZO1ByeXD4ZypUJwT96%2B2%2B%2BsQo%3D&product=jira&product=confluence

Known limitation, stated honestly

On the demo site, the Confluence token service could not issue a token for the app, so Confluence pages could not be read there. For the demo person this changes nothing, because she has no Confluence pages. The app reports the gap in the pack notes and the agent says the pack may be incomplete, instead of hiding it.

Built With

Share this project:

Updates

Submission history