jira-mcp

A Model Context Protocol (MCP) server that exposes Jira operations to AI assistants — search issues, create tickets, add comments, transition statuses, update fields, and attach files, all from within your AI chat.

Works with VS Code + GitHub Copilot, Claude Desktop, Cursor, and any MCP-compatible client.

Tools

| Tool | Description | |---|---| | jira_search | Search issues using JQL | | jira_get_issue | Get full details of an issue by key | | jira_create_issue | Create a new issue (Bug, Story, Task, etc.) | | jira_add_comment | Add a comment to an issue | | jira_get_my_issues | List open issues assigned to the current user | | jira_transition_issue | Move an issue to a new status | | jira_update_issue | Update fields (summary, priority, assignee, labels, team, epic link, etc.) | | jira_attach_file | Attach a local file to an issue |

Prerequisites

• Node.js 18 or later
• A Jira Personal Access Token (PAT)

Getting a Jira PAT

1. Log in to your Jira instance
2. Go to Profile → Personal Access Tokens (or <your-jira-url>/secure/ViewProfile.jspa)
3. Click Create token, give it a name, and copy the value

Installation

npm install -g @danielevilela/jira-mcp

Or use directly with npx — no install needed (see Configuration below).

Configuration

VS Code (GitHub Copilot Agent Mode)

Create or edit .vscode/mcp.json in your workspace:

{
  "servers": {
    "jira-mcp": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@danielevilela/jira-mcp"],
      "env": {
        "JIRA_URL": "https://your-company.atlassian.net",
        "JIRA_PAT": "your_personal_access_token"
      }
    }
  }
}

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{
  "mcpServers": {
    "jira-mcp": {
      "command": "npx",
      "args": ["-y", "@danielevilela/jira-mcp"],
      "env": {
        "JIRA_URL": "https://your-company.atlassian.net",
        "JIRA_PAT": "your_personal_access_token"
      }
    }
  }
}

Cursor

Add to Settings → MCP:

{
  "jira-mcp": {
    "command": "npx",
    "args": ["-y", "@danielevilela/jira-mcp"],
    "env": {
      "JIRA_URL": "https://your-company.atlassian.net",
      "JIRA_PAT": "your_personal_access_token"
    }
  }
}

Testing with MCP Inspector

The fastest way to manually test the server without an AI client:

JIRA_URL=https://your-company.atlassian.net \
JIRA_PAT=your_token \
npx @modelcontextprotocol/inspector npx @danielevilela/jira-mcp

This opens a browser UI where you can call each tool interactively.

Usage Examples

Once connected, you can ask your AI assistant things like:

"Search for open critical bugs in the DROP project"
"Create a story in DROP: [API] Add rate limiting to the auth endpoint"
"Transition DROP-1234 to In Progress"
"Add a comment to DROP-5678 saying the fix is deployed to staging"
"Show me all issues assigned to me"
"Update DROP-9999 priority to High and assign it to john.doe"

Environment Variables

| Variable | Required | Description | |---|---|---| | JIRA_URL | ✅ | Base URL of your Jira instance (e.g. https://company.atlassian.net) | | JIRA_PAT | ✅ | Personal Access Token for authentication |

Privacy & Data Storage

This MCP server is completely stateless — it stores nothing. Every tool call makes a live request directly to your Jira instance and returns the result to the AI. No ticket data, credentials, or history is written to disk or cached anywhere.

AI assistant → MCP tool call → Jira REST API → response back to AI

Your JIRA_PAT is passed only as an environment variable and is never logged or transmitted anywhere other than the Jira API.

License

MIT