The Jira MCP server gives Claude full read and write access to your Atlassian Jira workspace. For teams running their work in Jira, this is one of the most useful installs in the directory: ticket triage, sprint planning, and reporting all become single prompts.
The Atlassian-shipped server is the recommended path for most users. Connect via OAuth from the integrations panel in Claude Desktop and grant scoped access. No tokens to manage, easy to revoke.
Why use it
Jira’s UI is dense by design. It’s optimised for power users running structured agile processes. The flip side is that simple questions (“what’s blocked this sprint?”) often require navigating multiple boards, applying filters, and reading through tickets manually.
The MCP server flips this. Claude reads JQL, queries the API, returns a synthesised answer. Status updates, sprint reports, and dependency analysis become natural-language prompts.
What it actually does
A wide surface: search issues with JQL, fetch issue details with comments and attachments, list sprints and their issues, query users and teams, pull board configurations. With write scopes: create issues, transition statuses, add comments, log work, set fields.
Practical patterns:
- “What’s blocked this sprint? Pull every issue in ‘Sprint 24’ with status Blocked and show me the blockers.”
- “Create a Jira issue in PROJ titled ‘fix sitemap lastmod bug’, priority high, assigned to me.”
- “Summarise what shipped in the last release. Pull issues with the ‘fixed in 2.5’ label.”
Gotchas
Custom fields are everywhere in Jira. The server returns them but Claude may not know what they mean. If your workflow depends heavily on custom fields, give Claude a brief description in your prompt or system message.
Permissions are project-scoped. The OAuth flow grants access to all projects you can see, which may be more than you intend Claude to touch. For sensitive projects, consider running a separate Atlassian user account with limited project membership for the MCP integration.
Pair with GitHub and Sentry for end-to-end work tracking: Sentry catches the bug, Jira tracks it, GitHub holds the fix PR. Claude can move between all three without you switching tools.
For solo founders or small teams running Linear instead of Jira, Linear’s MCP server is generally lighter to install and gives a similar set of capabilities.