Jira

You can access Jira data through this MCP server, which provides relationship tracking, optimized payloads for AI context windows, and data cleaning to keep context concise. It supports both Jira Cloud and Jira Server/Data Center, making it practical for AI-assisted workflows that need Jira information, hierarchy, and history without bloating prompts.

How to use

Connect your MCP client to the Jira MCP Server using either the HTTP transport or the STDIO transport. The HTTP transport exposes a streamable endpoint, while STDIO runs locally and communicates over your editor or tool suite. You can perform tasks like searching issues, retrieving epic children with their history, getting detailed issue data, creating and updating issues, and adding comments or attachments. The server also cleans and transforms rich Jira content to suit AI context windows.

Typical use patterns include searching for issues with a JQL string, fetching epic children with comments, and then querying or modifying individual issues as part of your automation or AI-assisted workflows. All responses are optimized to minimize payloads while preserving essential relationship data such as mentions, links, parent/child relationships, and epics.

How to install

Prerequisites: you need Bun (v1.0.0 or higher) and a Jira account with API access.

Step 1: Prepare the environment variables and configuration for either STDIO or HTTP transport.

Step 2: Build and run the server as described below.

HTTP transport configuration

To enable the Streamable HTTP Transport, set the environment variables and start the server in HTTP mode. The HTTP endpoint is designed for MCP Inspector and other HTTP clients.

Example configuration and start command are shown here for reference. Use these exact commands and values to set up the HTTP transport and start the server.

export JIRA_API_TOKEN=your_api_token
export JIRA_BASE_URL=your_jira_instance_url
export JIRA_USER_EMAIL=your_email
export TRANSPORT_MODE=http
export HTTP_PORT=3000

bun run build/index.js

STDIO transport configuration

For the STDIO transport, configure a local MCP server that runs in your editor or Claude Desktop/Cline environment. This mode is ideal for production AI assistants and local development.

The complete command to run the local server is provided in the configuration snippet below. You will run a node process that starts the MCP server from the built index.

{
  "mcpServers": {
    "jira": {
      "command": "node",
      "args": ["/absolute/path/to/jira-mcp/build/index.js"],
      "env": {
        "JIRA_API_TOKEN": "your_api_token",
        "JIRA_BASE_URL": "your_jira_instance_url",
        "JIRA_USER_EMAIL": "your_email",
        "JIRA_TYPE": "cloud",
        "JIRA_AUTH_TYPE": "basic"
      }
    }
  }
}

Security and authentication

The server supports multiple authentication methods depending on Jira type. Jira Cloud uses API tokens with Basic authentication. Jira Server/Data Center supports Basic Auth or Bearer authentication with Personal Access Tokens (PATs). Always protect your API tokens and PATs, and prefer the bearer method for long-lived tokens if supported by your Jira instance.

Notes and limitations

Maximum limits: 50 issues per search results and 100 issues per epic children request. The server optimizes payloads for AI context windows and preserves essential relation data, including mentions, links, and epic associations.

Available tools

search_issues

Search Jira issues using JQL and return up to 50 results per request.

get_epic_children

Retrieve all child issues in an epic, including comments and relationship data, limited to 100 per request.

get_issue

Get detailed information about a specific Jira issue, including comments and related issues.

create_issue

Create a new Jira issue with specified fields.

update_issue

Update fields of an existing Jira issue.

add_attachment

Add a file attachment to a Jira issue.

add_comment

Add a comment to a Jira issue, with internal conversion to Atlassian Document Format.