Technical Guide: Connecting Jira to Your IDE via Atlassian MCP

The Model Context Protocol (MCP) provides a standardized interface for LLMs to interact with external data sources. Instead of writing custom integration code for every IDE, the Atlassian MCP server creates a secure, uniform bridge to your Jira instance.

This guide covers the technical implementation for connecting your development environment to Jira using the official Atlassian MCP server.

1. Architectural Overview

The connection utilizes a client-server architecture where your IDE (the MCP Host) communicates with the Atlassian MCP Server via Server-Sent Events (SSE).

• MCP Host: Your AI-enabled environment (e.g., Claude Desktop, Cursor, or VS Code).
• MCP Server: The mcp-remote proxy that handles the logic and authentication.
• Transport Layer: Standard HTTP with SSE for real-time tool execution.

2. Prerequisites

Ensure your local environment meets the following technical requirements:

• Runtime: Node.js (v18.0.0 or higher).
• Network: Outbound access to https://mcp.atlassian.com via port 443.
• Authentication: Valid Atlassian Cloud credentials with API access enabled for your user profile.

3. Configuration & Deployment

Open your MCP config file, usually named mcp_config.json or claude_desktop_config.json

Add the following JSON block to the mcpServers object:

"mcp-atlassian-api": {
  "command": "npx",
  "args": [
    "-y",
    "mcp-remote",
    "https://mcp.atlassian.com/v1/sse"
  ],
  "disabled": false
}

4. The Authentication Handshake

Atlassian MCP uses Three-Legged OAuth (3LO). Unlike legacy integrations, this does not require local storage of API tokens.

1. Trigger: Restart your MCP host. The mcp-remote process will initiate.
2. Redirect: A system browser window will open to id.atlassian.com.
3. Authorization: Upon successful login and “Allow” click, a short-lived authorization code is exchanged for an access token held in memory by the local proxy.

5. Verifying the Connection

To ensure the bridge is fully operational, follow these three verification steps:

A. Tool Discovery Check

In your AI assistant’s interface (e.g., the “Features” or “MCP” tab in Claude Desktop/Cursor), look for a list of registered tools. You should see a set of Atlassian-prefixed tools, such as:

• jira_search_issues
• jira_get_issue
• jira_edit_issue

B. Functional Query Test

Ask your AI assistant a specific, data-dependent question to verify read/write access:

“List the keys of the last 3 tickets updated in the [PROJECT_KEY] project.”

If it returns specific keys (e.g., DEV-101, DEV-102), the OAuth token and SSE stream are active.

C. Log Inspection

If the tools do not appear, inspect the MCP logs. For Claude Desktop, these are located at:

• macOS: ~/Library/Logs/Claude/mcp.log
• Windows: %APPDATA%\Claude\logs\mcp.log

Success Log Signature: [INFO] mcp-atlassian-api: connected via SSE

6. Troubleshooting Common Failures

• SSE Timeout: Ensure your corporate VPN or firewall isn’t blocking long-lived HTTP connections to mcp.atlassian.com.
• Node Version Mismatch: Run node -v. If it is below v18, the mcp-remote package will fail to initialize.
• Token Expiry: If the assistant suddenly loses access, restart the host to re-trigger the OAuth browser handshake.