@Atlassian@Claude @ChatGPTapp @Cursor In the end we switched for efficiency, we now store all new tasks in git issues on the respective repo the work should be carried out in and then assign it out to individuals so they simply login to gitea and know what they should be working on if they click issues/prs/milestones
@Atlassian@Claude @ChatGPTapp @Cursor We've left jira and confluence behind as a result of how inefficient this mcp tool is at providing agents the necessary means to manage without needing a human in the loop. And of the self hosted alternatives vikunja and taรฏga worked great, our agents could simply make api calls
@Atlassian@Claude @ChatGPTapp @Cursor Figuring out the mcp formatted calls and creating skills for each agent to load didn't improve usability as they'd still make numerous attempts to do a single edit or a single change to an item
@Atlassian@Claude @ChatGPTapp @Cursor We spent more time fixing our tooling to interface with jira and confluence than actually making sure the docs and tasks were aligned and accurate
@Atlassian@Claude @ChatGPTapp @Cursor Context window was toast the minute we made a an mcp call and then cursor would crash and fail to reopen the same workspace due to chat history being mangled and forcing a new workspace or a sqlite db surgical fix for the broken atlassian chat to remove it so it can start
@Atlassian@Claude @ChatGPTapp @Cursor Confluence it was impossible to edit a document using the mcp, we switched to using atlassian api from git to publish the MD formatted for confluence. Generating mermaid diagrams and publishing the images and embedding them correctly.