Project Status Reports with MCP: A Step-by-Step Playbook

A project status report tells stakeholders what shipped, what slipped and what is blocked. A project manager usually builds it every week by filtering Jira or Asana, reading comments and writing a summary by hand. With MCP (Model Context Protocol), an AI app can run those searches and draft the report for you, and the whole job can stay read-only until you choose to publish.

This playbook uses the Atlassian Rovo MCP server for Jira (and optionally Confluence) and the Asana MCP server. Both are official and hosted by the vendor. Both were checked against vendor docs on 28 October 2026. eSEOspace did not test them.

What you need

• An MCP client. Claude, ChatGPT, Cursor and Claude Code all appear in the vendors' setup guides.
• Jira: the Atlassian Rovo MCP server at https://mcp.atlassian.com/v2/mcp. Atlassian's getting started page gives that URL and says existing v1 use will switch to v2 tools on March 1, 2027.
• Asana: the Asana MCP server at https://mcp.asana.com/v2/mcp, per Asana's MCP docs.
• A report template. Five headings are enough: summary, done, in progress, blocked, next week.

For a step-by-step setup, see connect Claude to Atlassian.

The Jira tools that matter

Atlassian's supported tools page groups tools by permission. For a status report you only need these:

You do not need write_jira. Atlassian also says delete_jira and manage_jira are disabled by default.

The workflow

1. Fix the scope and dates. Name the project, board or Asana project and the reporting window. Ambiguity here produces a vague report.
2. Pull the work. In Jira, the AI calls searchJiraIssuesUsingJql with a query such as "project = WEB AND updated >= -7d." In Asana, ask for the project's status and its incomplete tasks due this week, which mirrors Asana's example prompts.
3. Get sprint or release context. listJiraBoardSprints and getJiraBoardSprintData give the current sprint. getJiraProjectVersions shows release versions if you plan by release.
4. Find what changed. listJiraIssueChangelogs shows status changes. This is how the AI tells "moved to Done this week" apart from "was already done."
5. Read the blockers. listJiraIssueComments and getJiraIssue pull the detail behind flagged or overdue items.
6. Draft the report. Ask for your five headings, with an issue key or task link next to every claim.
7. Check it. A person spot-checks the blocked list and every date. Then they add the judgment calls: risks, asks and tone.
8. Publish. Paste it into email or Slack yourself, or approve one createConfluenceContent call to save it as a Confluence page.

Example prompts

• "Search Jira for issues in project WEB updated in the last 7 days. Group them by status and list the issue keys. Do not change anything."
• "For the active sprint on the WEB board, list what was planned, what is done and what is still open."
• "Using the changelogs, list issues that moved to Done between 20 and 27 October."
• "In Asana, show me the status of the Client Portal project and all incomplete tasks due this week."
• "Draft a status report with these headings: Summary, Done, In progress, Blocked, Next week. Put the issue key after every line."

Guardrails

• Grant read and search only. Atlassian's README says organization admins grant or revoke access at the permission-group level, so read_jira and search_jira can be on while write_jira is off. If you publish to Confluence, add write_confluence for that step only.
• Check the audit log. The same README says every tool use through the server is recorded in your organization's audit log, under Insights, then Audit log, in Atlassian Administration.
• Mind the credits. Atlassian says credit usage is per call and depends on data volume and reasoning. Narrow JQL keeps both the report and the cost small.
• Treat comments as untrusted text. Comments are written by many people. A hostile line can try to steer the AI. Read our guide to MCP prompt injection.
• Keep a human in the loop. The MCP tools specification says clients should prompt for confirmation on sensitive operations and let a person deny tool calls.
• Know who can see what. Both servers act as the signed-in user. A director's account may surface confidential projects in a report meant for a client.
• Keep the narrative human. Risk calls, budget news and anything sent outside the company should be written or approved by the project lead.

More controls are in the MCP security hub.

What to measure

• Time to produce the weekly report, before and after.
• Errors found in review: wrong status, missed blocker, wrong date.
• How often stakeholders ask follow-up questions the report should have answered.
• Rovo credit use per report, if you are on Jira.
• Whether the report goes out on the same day each week.

Do not judge success by report length. A shorter, accurate report is the goal.

Next step

Find related task guides in the MCP playbooks hub. If you want this report automated across several tools with the right permissions, eSEOspace offers MCP development for custom servers and workflows.

Frequently asked questions