{"id":14867,"date":"2026-10-12T13:39:40","date_gmt":"2026-10-12T11:39:40","guid":{"rendered":"https:\/\/actonic.de\/?p=14867"},"modified":"2026-10-12T13:53:47","modified_gmt":"2026-10-12T11:53:47","slug":"jira-reports-mcp-server","status":"publish","type":"post","link":"https:\/\/actonic.de\/en\/jira-reports-mcp-server\/","title":{"rendered":"Ask Claude, Cursor or Codex for a Jira report: the Report Builder MCP server"},"content":{"rendered":"

[vc_row][vc_column][heading title=”The problem: your AI knows code, not your Jira reports” heading=”h2″][vc_column_text css=””]Many teams already spend the day in an AI client: Claude for writing and analysis, Cursor or Codex for code. But when someone asks “how many hours did the team log on Nova Commerce last month?” or “which bugs are still open, by priority?”, the answer lives in Jira. So you leave the chat, open Jira, find the right report or build one, pick fields, set filters, and come back.<\/p>\n

The AI could do the building, if it could reach your reporting tool safely and see only what you are allowed to see.[\/vc_column_text][heading title=”The solution: Report Builder has its own MCP server” heading=”h2″][vc_column_text css=””]Report Builder<\/a> now ships an MCP server. MCP (Model Context Protocol) is the open standard that lets AI clients use outside tools. Connect it once, and you can ask Claude, Cursor, Codex or any other MCP client for a Jira report in plain words. The AI finds the right Jira fields, builds the report and saves it in Report Builder, where it runs on live Jira data like any other report.<\/p>\n

Setup takes about a minute and has two parts: a personal token for your AI client and a Jira connection that lets Report Builder read Jira as you.[\/vc_column_text][heading title=”Watch: from setup to a saved report” heading=”h2″][vc_video link=”https:\/\/www.youtube.com\/watch?v=kSAES_rOu2Q” css=””][vc_column_text css=””]The video walks through the whole flow: issuing the token, adding it to an AI client, connecting Jira and asking for a first report.[\/vc_column_text][heading title=”How it works” heading=”h2″][vc_single_image image=”14857″ img_size=”full” alignment=”center” onclick=”link_image” css=””][vc_column_text css=””]There are two links. Your AI client talks to Report Builder with your personal MCP token. Report Builder reads Jira with your approval, given once in an Atlassian consent window. The report itself is built and saved in Report Builder, and the AI gives you a link to open it.[\/vc_column_text][heading title=”Set it up in about a minute” heading=”h2″][vc_column_text css=””]You need Report Builder on your Jira Cloud site with an active license, and an AI client that supports MCP. If your client starts the connection with npx<\/code> (Claude Desktop does), Node.js must be installed on your computer.<\/p>\n

1. Open Configure AI MCP.<\/strong> In Jira, open Report Builder, click the menu icon in the top-left corner and choose Configure AI MCP<\/strong>.<\/p>\n

2. Issue your MCP token.<\/strong> Click Issue MCP token<\/strong> and copy it right away. It is shown only once and is not stored. Lost it? Click Re-new MCP token<\/strong> for a fresh one.[\/vc_column_text][vc_single_image image=”14859″ img_size=”full” alignment=”center” onclick=”link_image” css=””][vc_column_text css=””]3. Pick your AI client.<\/strong> The same dialog has ready-made snippets for Claude Code, Claude Desktop, Cursor, Codex, Antigravity, OpenCode, OpenClaw and Hermes, with your token already filled in. Any other MCP client works with the same server address and X-MCP-API-Key<\/code> header.<\/p>\n

4. Paste the snippet into your client’s MCP configuration<\/strong>, save it and restart the client. In Claude Desktop that is claude_desktop_config.json<\/code>, under mcpServers<\/code>.<\/p>\n

5. Connect Jira.<\/strong> Click Connect Jira<\/strong>, choose your Jira site in the Atlassian consent window, review the access and accept. The status changes to Jira connected<\/strong>.[\/vc_column_text][vc_single_image image=”14861″ img_size=”full” alignment=”center” onclick=”link_image” css=””][vc_column_text css=””]That’s it. The page now shows both parts done: MCP token issued and Jira connected. You can renew or revoke the token here at any time.[\/vc_column_text][vc_single_image image=”14863″ img_size=”full” alignment=”center” onclick=”link_image” css=””][heading title=”Ask for a report in plain words” heading=”h2″][vc_column_text css=””]Open your AI client and describe what you want to see, for example:<\/p>\n

A bar chart of hours logged per teammate in Nova Commerce<\/p><\/blockquote>\n

The AI first checks whether a matching report already exists. If not, it builds a draft from a gallery template or from scratch. Open the draft from the link in the chat or from the Your report is ready<\/strong> notification in Report Builder. The report is not drawn in your chat; it lives in Report Builder.[\/vc_column_text][vc_single_image image=”14865″ img_size=”full” alignment=”center” onclick=”link_image” css=””][vc_column_text css=””]Want changes? Stay in the same chat: “make it a pie chart”, “only last month”. The AI updates the same draft. If something is unclear, it asks one question or tells you what it assumed. When you are happy, ask it to save the report or click Add to My Reports<\/strong>. Drafts expire after 24 hours, so save the ones you want to keep.<\/p>\n

What you can ask for<\/h3>\n\n\n\n\n\n\n\n\n\n
Task<\/th>\nExample request<\/th>\n<\/tr>\n
Find saved reports<\/td>\n“Do we already have a workload report for Nova Commerce?”<\/td>\n<\/tr>\n
Build a Universal Visual Builder report: numbers, bar, pie and doughnut charts, pivot tables, issue lists<\/td>\n“Open bugs by priority and assignee as a pivot table”<\/td>\n<\/tr>\n
Start from a gallery template: timesheets, workload, status distribution, calendar reports, Mermaid diagrams and more<\/td>\n“A Gantt chart of the epics in Nova Commerce”<\/td>\n<\/tr>\n
Build a custom scripted report (needs the Scripted Reports Creator permission)<\/td>\n“A one-page summary with total hours, finished issues and a weekly hours chart”<\/td>\n<\/tr>\n
Change an existing report, keeping its name and permissions<\/td>\n“Sort my Workload report by remaining estimate”<\/td>\n<\/tr>\n
Save a draft<\/td>\n“Save it as ‘Nova \u2014 hours per teammate'”<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

When you ask to change a saved report, the AI tells you whether it will update that report (if you can edit it) or create a copy. It makes a copy only when you ask for one.[\/vc_column_text][heading title=”Is it safe to let an AI into Jira?” heading=”h2″][vc_column_text css=””]This is usually the first question, so here is how it works:<\/p>\n