confluence

This MCP server provides a Confluence integration for AI assistants via a standardized MCP interface. It enables you to perform CQL searches, fetch full page content (with automatic table parsing), and update pages, all exposed through well-defined MCP actions. The server also includes smart data processing features such as automatic HTML table extraction and HTML tag cleanup to produce cleaner JSON representations of Confluence content. Authentication can be done using a Personal Access Token or via an email/API key pair, giving you flexibility to fit your security policies.

To use the tools, configure the MCP server in your environment with the appropriate CONFLUENCE_URL and credentials. The available actions include execute_cql_search, get_page_content, and update_page_content. For example, you can run a CQL search to locate pages by space and title, fetch a page’s content along with parsed tables, and then push updates to a page by providing new HTML content. The get_page_content response includes a tables array when the target page contains HTML tables, with headers and rows mapped into JSON objects for easy downstream consumption.

The server is designed to work with typical Confluence Cloud setups and exposes a clean data model for page metadata (id, title, space, version, url) as well as structured table data when present. This makes it suitable for AI workflows that need to search, summarize, or update Confluence content programmatically.

Prerequisites:

• Node.js 16+ (preferably LTS)
• npm or yarn
• Access to a Confluence instance with appropriate permissions

Install steps:

1. Clone or download the repository for the Confluence MCP Server.

git clone <repository-url>
cd confluence-mcp-server

2. Install dependencies.

npm install

3. Build the TypeScript sources to generate the runtime JS.

npm run build

4. Prepare environment variables. Create a .env file or export the variables in your environment. Example:

# Confluence instance URL
CONFLUENCE_URL=https://your-domain.atlassian.net/wiki

# Personal Access Token (recommended) or Email + API Key
CONFLUENCE_PERSONAL_TOKEN=your_personal_token_here
# Alternatively
# CONFLUENCE_API_MAIL=your-email@example.com
# CONFLUENCE_API_KEY=your_api_key_here

5. Run the MCP server via the MCP configuration. If you are using Claude Desktop or another MCP client, ensure the server entry is registered with the proper command and environment as shown in the mcp_config example. Example command used by the MCP server wrapper:

node build/index.js

6. Verify the server is reachable from your MCP client and perform a test call (e.g., execute_cql_search) to confirm connectivity.

Tips and common considerations:

• Ensure CONFLUENCE_URL uses the correct Cloud domain for your instance (e.g., https://your-domain.atlassian.net/wiki).
• Personal Tokens are the recommended authentication method for simplicity and security; API Keys are available as an alternative.
• If you encounter authentication failures, double-check token validity and permissions for the target Confluence space/pages.
• When using get_page_content, the response may include a tables array with parsed header rows and data rows. If a page contains complex or nested tables, some rows may require manual adjustments after parsing.
• The update_page_content tool accepts HTML content; ensure the content is properly escaped or constructed to avoid breaking the page formatting.
• For Claude Desktop integration, make sure the path to the built index.js matches your deployment environment (Windows vs. macOS/Linux) and update the env block accordingly.
• Monitor the MCP server logs for build or runtime errors during development and testing.