Note: This is a community-maintained project and is not affiliated with, endorsed by, or supported by Atlassian. Use at your own discretion.
This project provides a Model Context Protocol (MCP) integration for Atlassian Data Center products, including Jira, Confluence, and Bitbucket.
Each package ships an interactive setup subcommand that stores your credentials in the most secure place available on your OS. Run it once per product:
npx @atlassian-dc-mcp/jira setup
npx @atlassian-dc-mcp/confluence setup
npx @atlassian-dc-mcp/bitbucket setupThe setup CLI prompts for host, API base path, default page size, and API token. Before saving, it validates obvious input mistakes and performs a timed authenticated request to the selected Atlassian product, so a bad host, base path, or token is caught during setup.
Setup accepts flags so you can prefill values or skip prompts entirely (useful for scripted bootstrap, CI, or remote sessions). Run npx @atlassian-dc-mcp/<product> setup --help for the full list.
| Flag | Short | Description |
|---|---|---|
--host <value> |
-H |
Host, e.g. jira.example.com |
--api-base-path <value> |
-b |
API base path or full URL |
--token <value> |
-t |
API token |
--default-page-size <n> |
-s |
Default page size (positive integer) |
--non-interactive |
-n |
Skip prompts; fail if a required value cannot be resolved |
--help |
-h |
Show usage and exit |
In interactive mode, any flag you pass prefills its prompt (so e.g. --host skips the host prompt but still asks for the rest). In --non-interactive mode, setup resolves anything missing from existing configuration (process env, ~/.atlassian-dc-mcp/<product>.env, or macOS Keychain) and exits non-zero if a host (or full-URL --api-base-path) and token cannot be found. An existing token is reused when --token is omitted.
# Scripted, no prompts, write everything from flags
npx @atlassian-dc-mcp/jira setup --non-interactive \
--host jira.example.com \
--token "$JIRA_TOKEN"
# Re-validate the existing token without re-entering it
npx @atlassian-dc-mcp/jira setup --non-interactive --host jira.example.comCredential validation behaves differently between modes: interactive mode offers retry/save-anyway prompts on failure, while --non-interactive exits with code 1 on the first validation failure so it can be used as a CI gate.
Token storage:
- macOS — written to the login Keychain via
/usr/bin/security(serviceatlassian-dc-mcp, account<product>-token). - Linux — written to
~/.atlassian-dc-mcp/<product>.envwith POSIX mode0600(read/write for your user only; other local user accounts cannot read it). - Windows — written to
%USERPROFILE%\.atlassian-dc-mcp\<product>.env. Node passes the mode bits but Windows ignores them, so the file inherits the ACL of your user profile directory — typically readable only by your user, SYSTEM, and Administrators.
Non-secret fields (host, API base path, default page size) are always written to the home file — ~/.atlassian-dc-mcp/<product>.env on macOS/Linux, %USERPROFILE%\.atlassian-dc-mcp\<product>.env on Windows. After a successful Keychain write, the token line is cleared from the home file so there is never a second copy in a less-secure place.
Once setup has run, the MCP servers can boot with no environment variables at all:
{
"mcpServers": {
"atlassian-jira-dc": { "command": "npx", "args": ["-y", "@atlassian-dc-mcp/jira"] },
"atlassian-confluence-dc": { "command": "npx", "args": ["-y", "@atlassian-dc-mcp/confluence"] },
"atlassian-bitbucket-dc": { "command": "npx", "args": ["-y", "@atlassian-dc-mcp/bitbucket"] }
}
}You can still pass credentials via environment variables or ATLASSIAN_DC_MCP_CONFIG_FILE as shown in the sections below — they take precedence over values saved by setup.
At startup, each MCP server resolves each config key by walking sources in this order and taking the first non-empty value:
| Priority | Source | Reads | Written by setup |
|---|---|---|---|
| 100 | process.env (JIRA_*, CONFLUENCE_*, BITBUCKET_*) |
all keys | — |
| 80 | env file — ATLASSIAN_DC_MCP_CONFIG_FILE or ./.env |
all keys | — |
| 60 | home file — ~/.atlassian-dc-mcp/<product>.env on macOS/Linux, %USERPROFILE%\.atlassian-dc-mcp\<product>.env on Windows (mode 0600 on POSIX; Windows inherits the user-profile ACL) |
all keys | host, apiBasePath, defaultPageSize (always); token (non-darwin or keychain fallback) |
| 40 | macOS Keychain — service atlassian-dc-mcp, account <product>-token |
token only | token (darwin only) |
Notes:
- Process env wins over everything, so you can always override a stored credential for one session.
ATLASSIAN_DC_MCP_CONFIG_FILEmust be an absolute path; if set and missing, the server fails fast.- Keychain reads are cached at init (one
execFileSyncper product-token), so tool calls never shell out. - If a higher-priority source shadows the value setup is about to save, setup prints a warning naming the env var so you can unset it.
- Atlassian API requests time out after 30 seconds by default. Set
ATLASSIAN_DC_MCP_REQUEST_TIMEOUT_MSto a positive millisecond value to override it.
Official Anthropic quick start guide
To use these MCP connectors with Claude Desktop, add the following to your Claude Desktop configuration.
Set *_HOST variables only to domain + port without protocol (e.g., your-instance.atlassian.net). The https protocol is assumed.
Alternatively, you can use *_API_BASE_PATH variables instead of *_HOST to specify the complete API base URL including protocol (e.g., https://your-instance.atlassian.net/rest). Note that the /api/latest/ part is static and added automatically in the code, so you don't need to include it in the *_API_BASE_PATH values.
You can leave only the services you need in the configuration.
macOS:
~/Library/Application Support/Claude/claude_desktop_config.json
Windows:
%APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"atlassian-jira-dc": {
"command": "npx",
"args": ["-y", "@atlassian-dc-mcp/jira"],
"env": {
"JIRA_HOST": "your-jira-host",
"JIRA_API_TOKEN": "your-token"
}
},
"atlassian-confluence-dc": {
"command": "npx",
"args": ["-y", "@atlassian-dc-mcp/confluence"],
"env": {
"CONFLUENCE_HOST": "your-confluence-host",
"CONFLUENCE_API_TOKEN": "your-token"
}
},
"atlassian-bitbucket-dc": {
"command": "npx",
"args": ["-y", "@atlassian-dc-mcp/bitbucket"],
"env": {
"BITBUCKET_HOST": "your-bitbucket-host",
"BITBUCKET_API_TOKEN": "your-token"
}
}
}
}You can also use the alternative API base path configuration:
{
"mcpServers": {
"atlassian-jira-dc": {
"command": "npx",
"args": ["-y", "@atlassian-dc-mcp/jira"],
"env": {
"JIRA_API_BASE_PATH": "https://your-jira-host/rest",
"JIRA_API_TOKEN": "your-token"
}
},
"atlassian-confluence-dc": {
"command": "npx",
"args": ["-y", "@atlassian-dc-mcp/confluence"],
"env": {
"CONFLUENCE_API_BASE_PATH": "https://your-confluence-host/rest",
"CONFLUENCE_API_TOKEN": "your-token"
}
},
"atlassian-bitbucket-dc": {
"command": "npx",
"args": ["-y", "@atlassian-dc-mcp/bitbucket"],
"env": {
"BITBUCKET_API_BASE_PATH": "https://your-bitbucket-host/rest",
"BITBUCKET_API_TOKEN": "your-token"
}
}
}
}If you want multiple MCP hosts or tools on one machine to reuse the same Atlassian credentials, put the existing JIRA_*, CONFLUENCE_*, and BITBUCKET_* variables into one dotenv-style file and point each MCP server at it with ATLASSIAN_DC_MCP_CONFIG_FILE.
The path must be absolute. Direct environment variables still override values from the shared file.
Example shared file:
JIRA_HOST=your-jira-host
JIRA_API_TOKEN=your-jira-token
JIRA_DEFAULT_PAGE_SIZE=50
CONFLUENCE_HOST=your-confluence-host
CONFLUENCE_API_TOKEN=your-confluence-token
BITBUCKET_HOST=your-bitbucket-host
BITBUCKET_API_TOKEN=your-bitbucket-token
BITBUCKET_DEFAULT_PAGE_SIZE=50Claude Desktop example using one shared file:
{
"mcpServers": {
"atlassian-jira-dc": {
"command": "npx",
"args": ["-y", "@atlassian-dc-mcp/jira"],
"env": {
"ATLASSIAN_DC_MCP_CONFIG_FILE": "/Users/your-user/.config/atlassian-dc-mcp.env"
}
},
"atlassian-confluence-dc": {
"command": "npx",
"args": ["-y", "@atlassian-dc-mcp/confluence"],
"env": {
"ATLASSIAN_DC_MCP_CONFIG_FILE": "/Users/your-user/.config/atlassian-dc-mcp.env"
}
},
"atlassian-bitbucket-dc": {
"command": "npx",
"args": ["-y", "@atlassian-dc-mcp/bitbucket"],
"env": {
"ATLASSIAN_DC_MCP_CONFIG_FILE": "/Users/your-user/.config/atlassian-dc-mcp.env"
}
}
}
}Windows example path:
{
"mcpServers": {
"atlassian-jira-dc": {
"command": "npx",
"args": ["-y", "@atlassian-dc-mcp/jira"],
"env": {
"ATLASSIAN_DC_MCP_CONFIG_FILE": "C:\\\\Users\\\\your-user\\\\AppData\\\\Roaming\\\\atlassian-dc-mcp.env"
}
}
}
}To use these MCP connectors with Claude Code, add MCP servers using the claude mcp add command.
You can add servers at the project scope (stored in .mcp.json) or user scope (-s user). Adjust the scope and included services to your needs.
# Jira
claude mcp add atlassian-jira-dc \
-e JIRA_HOST=your-jira-host \
-e JIRA_API_TOKEN=your-token \
-- npx -y @atlassian-dc-mcp/jira
# Confluence
claude mcp add atlassian-confluence-dc \
-e CONFLUENCE_HOST=your-confluence-host \
-e CONFLUENCE_API_TOKEN=your-token \
-- npx -y @atlassian-dc-mcp/confluence
# Bitbucket
claude mcp add atlassian-bitbucket-dc \
-e BITBUCKET_HOST=your-bitbucket-host \
-e BITBUCKET_API_TOKEN=your-token \
-- npx -y @atlassian-dc-mcp/bitbucketYou can also use *_API_BASE_PATH instead of *_HOST (same as the Claude Desktop examples above):
claude mcp add atlassian-jira-dc \
-e JIRA_API_BASE_PATH=https://your-jira-host/rest \
-e JIRA_API_TOKEN=your-token \
-- npx -y @atlassian-dc-mcp/jiraTo add servers at user scope (available across all projects):
claude mcp add -s user atlassian-jira-dc \
-e JIRA_HOST=your-jira-host \
-e JIRA_API_TOKEN=your-token \
-- npx -y @atlassian-dc-mcp/jiraTo use the shared config file instead of passing credentials inline:
claude mcp add atlassian-jira-dc \
-e ATLASSIAN_DC_MCP_CONFIG_FILE=/Users/your-user/.config/atlassian-dc-mcp.env \
-- npx -y @atlassian-dc-mcp/jiraWindows PowerShell example:
claude mcp add atlassian-jira-dc `
-e ATLASSIAN_DC_MCP_CONFIG_FILE=C:\Users\your-user\AppData\Roaming\atlassian-dc-mcp.env `
-- npx -y @atlassian-dc-mcp/jiraFor Data Center installations, you'll need to generate Personal Access Tokens (PAT) for each service:
- Log in to your Jira instance
- Go to Profile > Personal Access Tokens
- Click "Create token"
- Give it a meaningful name and set appropriate permissions
- Copy the generated token immediately (it won't be shown again)
- Log in to your Confluence instance
- Go to Settings > Personal Access Tokens
- Click "Create token"
- Name your token and set required permissions
- Save and copy the token (only shown once)
- Log in to Bitbucket
- Go to Manage Account > HTTP access tokens
- Click "Create token"
- Set a name and permissions
- Generate and copy the token immediately
Store these tokens securely and use them in your Claude Desktop configuration as shown above.
The Atlassian DC MCP allows AI assistants to interact with Atlassian products through a standardized interface. It provides tools for:
- Jira: Search, view, and create issues
- Confluence: Access and manage content
- Bitbucket: Interact with repositories and code
- Node.js 18 or higher
- npm 7 or higher (for workspaces support)
- Atlassian Data Center instance or Cloud instance
- API tokens for the Atlassian products you want to use
Clone the repository:
git clone https://github.com/b1ff/atlassian-dc-mcp.git
cd atlassian-dc-mcpThis project is structured as an npm monorepo using workspaces. The workspaces are organized in the packages/ directory, with separate packages for each Atlassian product integration.
To install all dependencies for all packages in the monorepo:
npm installThis will install:
- Root-level dependencies defined in the root
package.json - All dependencies for each package in the workspaces
To install a dependency for a specific package:
npm install <package-name> --workspace=@atlassian-dc-mcp/jiraTo install a dependency at the root level:
npm install <package-name> -WTo build all packages:
npm run buildTo build a specific package:
npm run build --workspace=@atlassian-dc-mcp/jiraTo run a specific package in development mode:
npm run dev:jira # For Jira
npm run dev:confluence # For Confluence
npm run dev:bitbucket # For BitbucketFor production use, prefer the Quick Setup CLI above — it writes to the macOS Keychain (for tokens) and the home file for non-secret fields (~/.atlassian-dc-mcp/<product>.env on macOS/Linux, %USERPROFILE%\.atlassian-dc-mcp\<product>.env on Windows) automatically.
For local development, create a .env file in the root directory, or a shared dotenv-style file anywhere on disk and point ATLASSIAN_DC_MCP_CONFIG_FILE to it, with the following variables:
# Jira configuration - choose one of these options:
JIRA_HOST=your-instance.atlassian.net
# OR
JIRA_API_BASE_PATH=https://your-instance.atlassian.net/rest
# Note: part /api/2/search/ is added automatically, do not include it
JIRA_API_TOKEN=your-api-token
# Confluence configuration - choose one of these options:
CONFLUENCE_HOST=your-instance.atlassian.net
# OR
CONFLUENCE_API_BASE_PATH=https://your-instance.atlassian.net/confluence
# Note: part /rest/api is added automatically, do not include it
CONFLUENCE_API_TOKEN=your-api-token
# Bitbucket configuration - choose one of these options:
BITBUCKET_HOST=your-instance.atlassian.net
# OR
BITBUCKET_API_BASE_PATH=https://your-instance.atlassian.net/rest
# Note: part /api/latest/ is added automatically, do not include it
BITBUCKET_API_TOKEN=your-api-token
Resolution order for each key is process.env → ATLASSIAN_DC_MCP_CONFIG_FILE (or ./.env) → home file (~/.atlassian-dc-mcp/<product>.env on macOS/Linux, %USERPROFILE%\.atlassian-dc-mcp\<product>.env on Windows) → macOS Keychain. See Configuration Sources & Precedence above.
