jira-mcp
# 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**
## 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 (or username for Server/DC)
JIRA_TYPE=cloud # 'cloud' or 'server' (optional, defaults to 'cloud')
JIRA_AUTH_TYPE=basic # 'basic' or 'bearer' (optional, defaults to 'basic')
JIRA_SKIP_SSL_VERIFY=false # 'true' to skip SSL verification (optional, defaults to 'false')
```
### 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`
### SSL Certificate Verification
For Jira Server/Data Center instances using self-signed certificates, you can skip SSL verification:
```bash
JIRA_SKIP_SSL_VERIFY=true
```
> **Warning**: Only use this option in trusted environments. Disabling SSL verification makes connections vulnerable to man-in-the-middle attacks.
## 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
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": "https://your-jira-instance.com",
"JIRA_USER_EMAIL": "your_email_or_username",
"JIRA_TYPE": "cloud",
"JIRA_AUTH_TYPE": "basic",
"JIRA_SKIP_SSL_VERIFY": "false"
}
}
}
}
```
### 4. Restart the MCP server
Within Cline's MCP settings, restart the MCP server. Restart Claude Desktop to load the new MCP server.
## 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
- 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
## Error Handling
The server implements a comprehensive error handling strategy:
- Network error detection with helpful hints:
- SSL certificate errors (with suggestion to use `JIRA_SKIP_SSL_VERIFY`)
- Connection refused errors
- DNS lookup failures
- Connection timeout errors
- 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
## LICENCE
This project is licensed under the MIT License - see the [LICENCE](LICENCE) file for details.
TDQS
Scored across 9 tools
Each tool has a clearly distinct purpose: search vs. get specific issue vs. get epic children, get vs. perform transition, and add comment/attachment are well separated. Overlap is minimal and descriptions clarify boundaries.
All tool names follow a consistent snake_case verb_noun pattern (search_issues, get_issue, create_issue, etc.). The pattern is predictable throughout.
9 tools is a well-scoped set for JIRA issue management. Every tool earns its place, covering search, read, create, update, transitions, comments, and attachments without excess.
Core issue operations are present (search, get, create, update, transitions, comments, attachments), but lifecycle coverage is incomplete: no delete_issue, assign_issue, link_issue, or project/sprint operations. Agents may hit dead ends for deletion or assignment workflows.