# JIRA MCP Server
A Model Context Protocol (MCP) server implementation that provides access to JIRA data with relationship tracking, optimized data payloads, and data cleaning for AI context windows.
ℹ️ There is a separate MCP server [for Confluence](https://github.com/cosmix/confluence-mcp)
---
## Jira Cloud & Jira Server (Data Center) Support
This MCP server supports both **Jira Cloud** and **Jira Server (Data Center)** instances. You can select which type to use by setting the `JIRA_TYPE` environment variable:
- `cloud` (default): For Jira Cloud (Atlassian-hosted)
- `server`: For Jira Server/Data Center (self-hosted)
The server will automatically use the correct API version and authentication method for the selected type.
---
## Features
- Search JIRA issues using JQL (maximum 50 results per request)
- Retrieve epic children with comment history and optimized payloads (maximum 100 issues per request)
- Get detailed issue information including comments and related issues
- Create, update, and manage JIRA issues
- Add comments to issues
- Extract issue mentions from Atlassian Document Format
- Track issue relationships (mentions, links, parent/child, epics)
- Clean and transform rich JIRA content for AI context efficiency
- Support for file attachments with secure multipart upload handling
- **Supports both Jira Cloud and Jira Server (Data Center) APIs**
- **Dual transport modes: STDIO (default) and Streamable HTTP (MCP 2025-03-26) for tools like Postman and MCP Inspector**
## Prerequisites
- [Bun](https://bun.sh) (v1.0.0 or higher)
- JIRA account with API access
## Environment Variables
```bash
JIRA_API_TOKEN=your_api_token # API token for Cloud, PAT or password for Server/DC
JIRA_BASE_URL=your_jira_instance_url # e.g., https://your-domain.atlassian.net
JIRA_USER_EMAIL=your_email # Your Jira account email
JIRA_TYPE=cloud # 'cloud' or 'server' (optional, defaults to 'cloud')
JIRA_AUTH_TYPE=basic # 'basic' or 'bearer' (optional, defaults to 'basic')
TRANSPORT_MODE=stdio # 'stdio' or 'http' (optional, defaults to 'stdio')
HTTP_PORT=3000 # Port for HTTP transport (optional, defaults to 3000)
```
### Authentication Methods
- **Jira Cloud**: Use API tokens with Basic authentication
- Create an API token at: <https://id.atlassian.com/manage-profile/security/api-tokens>
- Set `JIRA_AUTH_TYPE=basic` (default)
- **Jira Server/Data Center**:
- **Basic Auth**: Use username/password or API tokens
- Set `JIRA_AUTH_TYPE=basic` (default)
- **Bearer Auth**: Use Personal Access Tokens (PATs) - available in Data Center 8.14.0+
- Create a PAT in your profile settings
- Set `JIRA_AUTH_TYPE=bearer`
- Use the PAT as your `JIRA_API_TOKEN`
## Installation & Setup
### 1. Clone the repository
```bash
git clone [repository-url]
cd jira-mcp
```
### 2. Install dependencies and build
```bash
bun install
bun run build
```
### 3. Configure the MCP server
The server supports two transport modes: **STDIO** (default, for Claude Desktop/Cline) and **HTTP** (for Postman and other HTTP-based clients).
#### Option A: STDIO Transport (Default - Claude Desktop/Cline)
Edit the appropriate configuration file:
**macOS:**
- Cline: `~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json`
- Claude Desktop: `~/Library/Application Support/Claude/claude_desktop_config.json`
**Windows:**
- Cline: `%APPDATA%\Code\User\globalStorage\saoudrizwan.claude-dev\settings\cline_mcp_settings.json`
- Claude Desktop: `%APPDATA%\Claude Desktop\claude_desktop_config.json`
**Linux:**
- Cline: `~/.config/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json`
- Claude Desktop: _sadly doesn't exist yet_
Add the following configuration under the `mcpServers` object:
```json
{
"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"
}
}
}
}
```
#### Option B: Streamable HTTP Transport (MCP Inspector & Postman)
1. **Start the server in HTTP mode:**
```bash
# Set environment variables
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" # optional, defaults to 3000
# Run the server
bun run build/index.js
```
Or use a `.env` file:
```bash
# .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
TRANSPORT_MODE=http
HTTP_PORT=3000
```
2. **The server will start with Streamable HTTP endpoint:**
```
╔════════════════════════════════════════════════════════════╗
║ 🚀 JIRA MCP Server - Streamable HTTP Transport ║
╠════════════════════════════════════════════════════════════╣
║ Protocol: MCP 2025-03-26 (Streamable HTTP) ║
║ Endpoint: http://localhost:3000/mcp ║
║ Health Check: http://localhost:3000/health ║
║ ║
║ 🔧 Supported Methods: ║
║ GET /mcp → Establish SSE connection ║
║ POST /mcp → Send JSON-RPC message ║
║ DELETE /mcp → Terminate session ║
╚════════════════════════════════════════════════════════════╝
```
3. **Configure MCP Inspector:**
In MCP Inspector v0.15.x+, add a new server connection:
```json
{
"mcpServers": {
"jira": {
"type": "sse",
"url": "http://localhost:3000/mcp"
}
}
}
```
4. **Configure Postman MCP Client:**
In Postman, add a new MCP server connection:
- **URL**: `http://localhost:3000/mcp`
- **Method**: GET
- **Transport**: SSE (Server-Sent Events)
Or import the provided Postman collection: `tests/manual/postman_collection.json`
The server implements the MCP Streamable HTTP protocol (2025-03-26) with full session management and is compatible with MCP Inspector v0.15.x and Postman MCP Client.
### 4. Restart the MCP server
**For STDIO mode**: Within Cline's MCP settings, restart the MCP server. Restart Claude Desktop to load the new MCP server.
**For HTTP mode**: Simply run the server with `TRANSPORT_MODE=http` environment variable set.
## Development
Run tests:
```bash
bun test
```
Watch mode for development:
```bash
bun run dev
```
To rebuild after changes:
```bash
bun run build
```
## Available MCP Tools
### search_issues
Search JIRA issues using JQL. Returns up to 50 results per request.
Input Schema:
```typescript
{
searchString: string; // JQL search string
}
```
### get_epic_children
Get all child issues in an epic including their comments and relationship data. Limited to 100 issues per request.
Input Schema:
```typescript
{
epicKey: string; // The key of the epic issue
}
```
### get_issue
Get detailed information about a specific JIRA issue including comments and all relationships.
Input Schema:
```typescript
{
issueId: string; // The ID or key of the JIRA issue
}
```
### create_issue
Create a new JIRA issue with specified fields.
Input Schema:
```typescript
{
projectKey: string, // The project key where the issue will be created
issueType: string, // The type of issue (e.g., "Bug", "Story", "Task")
summary: string, // The issue summary/title
description?: string, // Optional issue description
fields?: { // Optional additional fields
[key: string]: any
}
}
```
### update_issue
Update fields of an existing JIRA issue.
Input Schema:
```typescript
{
issueKey: string, // The key of the issue to update
fields: { // Fields to update
[key: string]: any
}
}
```
### add_attachment
Add a file attachment to a JIRA issue.
Input Schema:
```typescript
{
issueKey: string, // The key of the issue
fileContent: string, // Base64 encoded file content
filename: string // Name of the file to be attached
}
```
### add_comment
Add a comment to a JIRA issue. Accepts plain text and converts it to the required Atlassian Document Format internally.
Input Schema:
```typescript
{
issueIdOrKey: string, // The ID or key of the issue to add the comment to
body: string // The content of the comment (plain text)
}
```
## Data Cleaning Features
- Extracts text from Atlassian Document Format
- Tracks issue mentions in descriptions and comments
- Maintains formal issue links with relationship types
- Preserves parent/child relationships
- Tracks epic associations
- Includes comment history with author information
- Removes unnecessary metadata from responses
- Recursively processes content nodes for mentions
- Deduplicates issue mentions
## Technical Details
- Built with TypeScript in strict mode
- Uses Bun runtime for improved performance
- Vite for optimized builds
- **Dual transport support:**
- **STDIO**: For Claude Desktop and Cline (default)
- **Streamable HTTP (MCP 2025-03-26)**: For MCP Inspector v0.15.x+ and Postman MCP Client
- Session management with automatic cleanup (1-hour expiry)
- UUID-based session tracking with `Mcp-Session-Id` headers
- Uses JIRA REST API v3 (Cloud) or v2 (Server/Data Center)
- Supports multiple authentication methods:
- Basic authentication with API tokens or username/password
- Bearer authentication with Personal Access Tokens (PATs)
- Batched API requests for related data
- Optimized response payloads for AI context windows
- Efficient transformation of complex Atlassian structures
- Robust error handling
- Rate limiting considerations
- Maximum limits:
- Search results: 50 issues per request
- Epic children: 100 issues per request
- Support for multipart form data for secure file attachments
- Automatic content type detection and validation
- CORS enabled for cross-origin requests in HTTP mode
## Error Handling
The server implements a comprehensive error handling strategy:
- Network error detection and appropriate messaging
- HTTP status code handling (especially 404 for issues)
- Detailed error messages with status codes
- Error details logging to console
- Input validation for all parameters
- Safe error propagation through MCP protocol
- Specialized handling for common JIRA API errors
- Base64 validation for attachments
- Multipart request failure handling
- Rate limit detection
- Attachment parameter validation
## Transport Modes Comparison
| Feature | STDIO Transport | Streamable HTTP Transport |
|---------|----------------|---------------------------|
| **Use Case** | Claude Desktop, Cline | MCP Inspector, Postman, HTTP clients |
| **Protocol** | MCP over stdio | MCP Streamable HTTP (2025-03-26) |
| **Configuration** | MCP config file | Environment variables |
| **Connection** | Process stdin/stdout | HTTP + Server-Sent Events |
| **Default Port** | N/A | 3000 |
| **Session Management** | N/A | Yes (UUID-based, 1-hour expiry) |
| **Endpoints** | N/A | GET/POST/DELETE `/mcp` |
| **CORS Support** | N/A | Yes |
| **Health Check** | No | Yes (`/health` endpoint) |
| **Best For** | Production AI assistants | Development, testing, debugging |
| **Inspector Compatibility** | N/A | v0.15.x+ |
### Manual Testing
The `tests/manual/` directory contains comprehensive test scripts:
- `test_sse.sh` - Test SSE connection and session establishment
- `test_initialize.sh` - Test initialize handshake
- `test_tools_list.sh` - Test tools/list request
- `test_tool_execution.sh` - Test tool execution (search_issues)
- `postman_collection.json` - Complete Postman test collection
- `README.md` - Detailed testing instructions
Run all tests:
```bash
cd tests/manual
./test_sse.sh
./test_initialize.sh
./test_tools_list.sh
./test_tool_execution.sh
```
## LICENCE
This project is licensed under the MIT License - see the [LICENCE](LICENCE) file for details.
