@leilashellroot/confluence-data-center-mcp

MCP server for Confluence Server/Data Center. It gives AI assistants access to content, spaces, comments, attachments, metadata, and complete document context through the Confluence REST API.

This project is an independent community package and is not affiliated with Atlassian.

This project is based on and credits b1ff/atlassian-dc-mcp.

Features

• Content CRUD for pages, blog posts, and comments
• CQL content and space search
• Cursor-based content scanning
• Direct children, descendants, history, and historical versions
• Complete document context aggregation
• Root, nested, footer, inline, and resolved comment retrieval
• Attachments, labels, content properties, and restrictions
• Space listing, content listing, creation, updates, deletion, and archiving
• Watcher and user lookup
• stdio, legacy SSE, and Streamable HTTP transports
• Interactive and non-interactive secure setup

Installation

npm install -g @leilashellroot/confluence-data-center-mcp

The server requires a Confluence Server/Data Center personal access token or API token.

Configuration

The simplest configuration is:

export CONFLUENCE_BASE_URL=https://confluence.example.com
export CONFLUENCE_API_TOKEN=your-token

These equivalent names are also supported:

export CONFLUENCE_HOST=confluence.example.com
export CONFLUENCE_PAT=your-token

For an installation below a context path, use a full API base path:

export CONFLUENCE_API_BASE_PATH=https://confluence.example.com/wiki/rest/api
export CONFLUENCE_API_TOKEN=your-token

The server removes the final /rest or /rest/api suffix before constructing API requests.

Configuration precedence is process environment, ATLASSIAN_DC_MCP_CONFIG_FILE or the current directory .env, ~/.atlassian-dc-mcp/confluence.env, then the macOS Keychain token.

CONFLUENCE_DEFAULT_PAGE_SIZE defaults to 25. ATLASSIAN_DC_MCP_REQUEST_TIMEOUT_MS defaults to 30000.

Setup CLI

The setup command validates credentials against /rest/api/user/current and stores configuration in ~/.atlassian-dc-mcp/confluence.env with mode 0600 on POSIX systems:

npx @leilashellroot/confluence-data-center-mcp setup

Scripted setup:

npx @leilashellroot/confluence-data-center-mcp setup --non-interactive \
  --host confluence.example.com \
  --token "$CONFLUENCE_TOKEN"

Available flags are --host/-H, --api-base-path/-b, --token/-t, --default-page-size/-s, --non-interactive/-n, and --help/-h.

On macOS, setup stores the token in the login Keychain under the atlassian-dc-mcp service and confluence-token account. On other systems it stores the token in the protected home configuration file.

Usage

Direct stdio usage:

npx @leilashellroot/confluence-data-center-mcp

OpenCode, Cursor, or Claude Desktop:

{
  "mcp": {
    "confluence": {
      "type": "local",
      "command": ["npx", "-y", "@leilashellroot/confluence-data-center-mcp"],
      "environment": {
        "CONFLUENCE_BASE_URL": "https://confluence.example.com",
        "CONFLUENCE_API_TOKEN": "your-token"
      },
      "enabled": true
    }
  }
}

After setup, credentials can be omitted from the MCP configuration:

{
  "mcpServers": {
    "confluence": {
      "command": "npx",
      "args": ["-y", "@leilashellroot/confluence-data-center-mcp"]
    }
  }
}

Full Context

confluence_get_content_context is the primary agent-oriented tool. It returns:

• The complete content object and complete storage-format body
• A derived plain-text document body
• All comment pages across pagination boundaries
• Footer, inline, and resolved comments
• Nested replies assembled into comment trees
• Inline properties and resolution metadata
• Author, timestamps, IDs, parent IDs, and raw Confluence fields
• Counts and query metadata showing the aggregation was complete

The context tool does not truncate the document or comments. Large pages can produce large MCP responses, so use confluence_get_content with bodyMode set to text or none for targeted reads.

Tools

Content

confluence_get_content, confluence_get_content_context, confluence_search_content, confluence_scan_content, confluence_create_content, confluence_update_content, confluence_delete_content, confluence_get_children, confluence_get_descendants, confluence_get_history, confluence_get_content_version, and confluence_delete_content_version.

Comments

confluence_get_comments, confluence_get_all_comments, confluence_create_comment, confluence_update_comment, and confluence_delete_comment.

Comment bodies use Confluence storage format. Inline creation can pass inlineProperties when the target Data Center version supports the relevant inline-comment contract.

Spaces

confluence_get_spaces, confluence_search_spaces, confluence_get_space, confluence_get_space_content, confluence_create_space, confluence_update_space, confluence_delete_space, and confluence_archive_space.

Attachments and metadata

confluence_get_attachments, confluence_get_attachment, confluence_add_attachment, confluence_update_attachment, confluence_update_attachment_data, confluence_delete_attachment, confluence_get_labels, confluence_add_labels, confluence_remove_label, confluence_get_content_properties, confluence_get_content_property, confluence_create_content_property, confluence_update_content_property, and confluence_delete_content_property.

Permissions and users

confluence_get_restrictions, confluence_update_restrictions, confluence_get_watchers, confluence_get_current_user, confluence_get_user, and confluence_get_users.

The original reference tool names remain available as compatibility aliases: confluence_getContent, confluence_searchContent, confluence_createContent, confluence_updateContent, and confluence_searchSpace.

Remote Transports

Legacy SSE:

MCP_TRANSPORT=sse MCP_HOST=0.0.0.0 MCP_PORT=3000 \
  npx @leilashellroot/confluence-data-center-mcp

Streamable HTTP:

MCP_TRANSPORT=streamable-http MCP_HOST=0.0.0.0 MCP_PORT=3000 \
  npx @leilashellroot/confluence-data-center-mcp

Remote configuration variables:

| Variable | Default | Description | |---|---|---| | MCP_TRANSPORT | stdio | stdio, sse, or streamable-http | | MCP_HOST | 127.0.0.1 | HTTP bind host | | MCP_PORT | 3000 | HTTP port | | MCP_HTTP_PATH | /mcp | Streamable HTTP endpoint | | MCP_SSE_PATH | /sse | SSE endpoint | | MCP_MESSAGES_PATH | /messages | SSE message endpoint | | MCP_ALLOWED_HOSTS | unset | Comma-separated allowed Host values | | MCP_ALLOWED_ORIGINS | unset | Comma-separated allowed Origin values | | MCP_CORS_ORIGIN | unset | Optional CORS origin |

When binding beyond localhost, use HTTPS and put the server behind authentication. Every connected client uses the configured Confluence token.

Development

npm install
npm test
npm run build

The test suite uses mocked REST responses and local HTTP servers. It does not require access to a Confluence instance.

Publishing

The tag-triggered GitHub Actions workflow expects an npm access token in the repository secret named NPM_TOKEN. The token is supplied to npm through NODE_AUTH_TOKEN only during the publish step and is not stored in the repository.

API References

• Confluence Data Center REST API
• Confluence Server REST API reference
• Model Context Protocol SDK

License

MIT