Because search takes JQL rather than a bespoke filter language, whatever you already know about querying Jira transfers directly and the assistant can build precise queries instead of listing everything. Transitions are handled properly too — it lists what is available for that issue before moving it, rather than guessing at a status name. Auth is deliberately flexible, but note that the sample Claude Desktop config in the README is written for token auth only.
An MCP server over Jira's REST API built on the `jira-python` library, working against Jira Cloud, Server or Data Center. It covers the everyday loop: find issues, read one, create one, comment, and transition it.
- Every project you have access to, listed
- One issue's full detail by key
- Issues found with JQL, so the filters you already use in Jira work here
- New issues created, with project, summary, description and issue type
- Comments added to an issue
- Available transitions listed for an issue, and the issue moved to a new status
- Bulk creation converted to the v3 REST API — up to 50 issues per call, with string descriptions converted to Atlassian Document Format automatically and issue types case-normalised ("bug" becomes "Bug")
A Jira instance and credentials. Set `JIRA_SERVER_URL`, then `JIRA_AUTH_METHOD` to `basic_auth` or `token_auth`; basic auth uses `JIRA_USERNAME` and `JIRA_PASSWORD`, token auth uses `JIRA_USERNAME` and `JIRA_TOKEN`. Python 3.9 or higher; install with `uv pip install mcp-server-jira` or `pip install mcp-server-jira`, then run `python -m mcp_server_jira` or point the client at `uv --directory <path> run mcp-server-jira`. A Dockerfile is included. Package `mcp-server-jira` 0.1.0.
One command plus a key — pip install mcp-server-jira, then supply credentials