Jira MCP

What jira-mcp MCP server does

The jira-mcp MCP server exposes Jira operations to an MCP-compatible AI client. It is intended for teams that want to inspect and modify Jira issues through natural-language prompts while keeping the server on a local machine. Supported work includes listing and searching issues, creating tickets, updating existing tickets, changing issue status through transitions, and managing ticket data.

The project is written for Python 3.12 and is distributed under the MIT license. Release binaries are available for Linux, Windows, macOS on Apple Silicon, and macOS on Intel processors.

How it works

A client such as Claude Desktop, Cursor, Windsurf, or ChatGPT sends a user request to an LLM. The model selects an available Jira tool, and the client invokes that tool through the local MCP process. The server passes the operation to jira-cli, receives the result, and returns it to the client so the model can present the result or report the completed action.

The jira-mcp MCP server uses MCP tool descriptions to give the model a structured interface rather than requiring the model to construct terminal commands or Jira endpoints directly. Because the process runs locally, Jira credentials and requests are handled through the configured local environment and jira-cli installation.

Setup and configuration

Install jira-cli first, then run jira init to configure the Jira URL and credentials. Verify the CLI setup with jira issue list before adding the MCP server to a client.

The server requires these environment variables:

• JIRA_API_TOKEN: the Jira API token.
• JIRA_AUTH_TYPE: the authentication mode. The documented values are bearer, basic, and password.

The README recommends placing these values in the MCP client's env configuration. This is especially relevant for desktop applications that may not inherit variables from shell startup files.

For the simplest deployment, download the binary for the host operating system and register it as an MCP server with a command pointing to the executable. The repository also documents a development configuration using uv --directory ... run python -m src.main, plus an editable installation through pipx. macOS may require approving the downloaded executable in Privacy & Security or removing its quarantine attribute.

Tools and capabilities

The jira-mcp MCP server is designed for Jira issue workflows, including:

• Searching or listing Jira issues.
• Creating new tickets.
• Updating ticket fields.
• Transitioning issues between workflow states.
• Managing tickets through a conversational client interface.

Exact field support, project permissions, and workflow transitions therefore depend on the Jira account, jira-cli configuration, and Jira instance.

Limitations and notes

Jira access still depends on valid credentials, the configured Jira URL, jira-cli, and the permissions granted to the account. The server does not replace Jira permissions or guarantee that a requested transition exists in a project's workflow.

The process communicates with MCP clients using standard protocol messages. Application logging must not be written with print to standard output, because that can corrupt JSON-RPC traffic. Use the project's documented logging approach when modifying or developing the server.

The README names Claude Desktop, Cursor, Windsurf, and ChatGPT as example clients. Client-specific configuration is not included beyond the general MCP server JSON pattern, so the command, executable path, and environment values must be adapted to the selected application.