Contentful MCP Server

Notice

This is a community driven server! Contentful has released an official server which you can find here

An MCP server implementation that integrates with Contentful's Content Management API, providing comprehensive content management capabilities.

- Please note \*; if you are not interested in the code, and just want to use this MCP in
Claude Desktop (or any other tool that is able to use MCP servers) you don't have to
clone this repo, you can just set it up in Claude desktop, refer to the section
"Usage with Claude Desktop" for instructions on how to install it.

<a href="https://glama.ai/mcp/servers/l2fxeaot4p"></a>

Features

- Content Management: Full CRUD operations for entries and assets
- Comment Management: Create, retrieve, and manage comments on entries with support for both plain-text and rich-text formats, including threaded conversations
- Space Management: Create, update, and manage spaces and environments
- Content Types: Manage content type definitions
- Localization: Support for multiple locales
- Publishing: Control content publishing workflow
- Bulk Operations: Execute bulk publishing, unpublishing, and validation across multiple entries and assets
- Smart Pagination: List operations return maximum 3 items per request to prevent context window overflow, with built-in pagination support

Pagination

To prevent context window overflow in LLMs, list operations (like search_entries and list_assets) are limited to 3 items per request. Each response includes:

- Total number of available items
- Current page of items (max 3)
- Number of remaining items
- Skip value for the next page
- Message prompting the LLM to offer retrieving more items

This pagination system allows the LLM to efficiently handle large datasets while maintaining context window limits.

Bulk Operations

The bulk operations feature provides efficient management of multiple content items simultaneously:

- Asynchronous Processing: Operations run asynchronously and provide status updates
- Efficient Content Management: Process multiple entries or assets in a single API call
- Status Tracking: Monitor progress with success and failure counts
- Resource Optimization: Reduce API calls and improve performance for batch operations

These bulk operation tools are ideal for content migrations, mass updates, or batch publishing workflows.

Tools

Entry Management

- search_entries: Search for entries using query parameters
- create_entry: Create new entries
- get_entry: Retrieve existing entries
- update_entry: Update entry fields
- delete_entry: Remove entries
- publish_entry: Publish entries
- unpublish_entry: Unpublish entries

Comment Management

- get_comments: Retrieve comments for an entry with filtering by status (active, resolved, all)
- create_comment: Create new comments on entries with support for both plain-text and rich-text formats. Supports threaded conversations by providing a parent comment ID to reply to existing comments
- get_single_comment: Retrieve a specific comment by its ID for an entry
- delete_comment: Delete a specific comment from an entry
- update_comment: Update existing comments with new body content or status changes

Threaded Comments

Comments support threading functionality to enable structured conversations and work around the 512-character limit:

- Reply to Comments: Use the parent parameter in create_comment to reply to an existing comment
- Threaded Conversations: Build conversation trees by replying to specific comments
- Extended Discussions: Work around the 512-character limit by creating threaded replies to continue longer messages
- Conversation Context: Maintain context in discussions by organizing related comments in threads

Example usage:

1. Create a main comment: create_comment with entryId, body, and status
2. Reply to that comment: create_comment with entryId, body, status, and parent (the ID of the comment you're replying to)
3. Continue the thread: Reply to any comment in the thread by using its ID as the parent

Bulk Operations

- bulk_publish: Publish multiple entries and assets in a single operation. Accepts an array of entities (entries and assets) and processes their publication as a batch.
- bulk_unpublish: Unpublish multiple entries and assets in a single operation. Similar to bulk_publish but removes content from the delivery API.
- bulk_validate: Validate multiple entries for content consistency, references, and required fields. Returns validation results without modifying content.

Asset Management

- list_assets: List assets with pagination (3 items per page)
- upload_asset: Upload new assets with metadata
- get_asset: Retrieve asset details and information
- update_asset: Update asset metadata and files
- delete_asset: Remove assets from space
- publish_asset: Publish assets to delivery API
- unpublish_asset: Unpublish assets from delivery API

Space & Environment Management

- list_spaces: List available spaces
- get_space: Get space details
- list_environments: List environments in a space
- create_environment: Create new environment
- delete_environment: Remove environment

Content Type Management

- list_content_types: List available content types
- get_content_type: Get content type details
- create_content_type: Create new content type
- update_content_type: Update content type
- delete_content_type: Remove content type
- publish_content_type: Publish a content type

Development Tools

MCP Inspector

The project includes an MCP Inspector tool that helps with development and debugging:

- Inspect Mode: Run npm run inspect to start the inspector, you can open the inspector by going to http://localhost:5173
- Watch Mode: Use npm run inspect:watch to automatically restart the inspector when files change
- Visual Interface: The inspector provides a web interface to test and debug MCP tools
- Real-time Testing: Try out tools and see their responses immediately
- Bulk Operations Testing: Test and monitor bulk operations with visual feedback on progress and results

The project also contains a npm run dev command which rebuilds and reloads the MCP server on every change.

Configuration

Prerequisites

1. Create a Contentful account at Contentful
2. Generate a Content Management API token from your account settings

Environment Variables

These variables can also be set as arguments

- CONTENTFUL_HOST / --host: Contentful Management API Endpoint (defaults to https://api.contentful.com)
- CONTENTFUL_MANAGEMENT_ACCESS_TOKEN / --management-token: Your Content Management API token
- ENABLE_HTTP_SERVER / --http: Set to "true" to enable HTTP/SSE mode
- HTTP_PORT / --port: Port for HTTP server (default: 3000)
- HTTP_HOST / --http-host: Host for HTTP server (default: localhost)
- DISABLE_AI_ACTIONS: Set to "true" to disable fetching AI Actions on startup (useful if you don't have access to this feature)

Space and Environment Scoping

You can scope the spaceId and EnvironmentId to ensure the LLM will only do operations on the defined space/env ID's.
This is mainly to support agents that are to operate within specific spaces. If both SPACE_ID and ENVIRONMENT_ID env-vars are set
the tools will not report needing these values and the handlers will use the environment vars to do CMA operations.
You will also loose access to the tools in the space-handler, since these tools are across spaces.
You can also add the SPACE_ID and ENVIRONMENT_ID by using arguments --space-id and --environment-id

Using App Identity

Instead of providing a Management token you can also leverage App Identity
for handling authentication. You would have to setup and install a Contentful App and set the following parameters when calling the MCP-server:

- --app-id = the app Id which is providing the Apptoken
- --private-key = the private key you created in the user-interface with your app, tied to app_id
- --space-id = the spaceId in which the app is installed
- --environment-id = the environmentId (within the space) in which the app is installed.

With these values the MCP server will request a temporary AppToken to do content operation in the defined space/environment-id. This especially useful when using this MCP server in backend systems that act as MCP-client (like chat-agents)

Usage with Claude Desktop

You do not need to clone this repo to use this MCP, you can simply add it to
your claude_desktop_config.json:

Add or edit ~/Library/Application Support/Claude/claude_desktop_config.json
and add the following lines:

{
  "mcpServers": {
    "contentful": {
      "command": "npx",
      "args": ["-y", "@ivotoby/contentful-management-mcp-server"],
      "env": {
        "CONTENTFUL_MANAGEMENT_ACCESS_TOKEN": "<Your CMA token>"
      }
    }
  }
}

If your MCPClient does not support setting environment variables you can also set the management token using an argument like this:

{
  "mcpServers": {
    "contentful": {
      "command": "npx",
      "args": [
        "-y",
        "@ivotoby/contentful-management-mcp-server",
        "--management-token",
        "<your token>",
        "--host",
        "http://api.contentful.com"
      ]
    }
  }
}

Installing via Smithery

To install Contentful Management Server for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install @ivotoby/contentful-management-mcp-server --client claude

Developing and using Claude desktop

If you want to contribute and test what Claude does with your contributions;

- run npm run dev, this will start the watcher that rebuilds the MCP server on every change
- update claude_desktop_config.json to reference the project directly, ie;

{
  "mcpServers": {
    "contentful": {
      "command": "node",
      "args": ["/Users/ivo/workspace/contentful-mcp/bin/mcp-server.js"],
      "env": {
        "CONTENTFUL_MANAGEMENT_ACCESS_TOKEN": "<Your CMA Token>"
      }
    }
  }
}

This will allow you to test any modification in the MCP server with Claude directly, however; if you add new tools/resources you will need to restart Claude Desktop

Transport Modes

The MCP server supports two transport modes:

stdio Transport

The default transport mode uses standard input/output streams for communication. This is ideal for integration with MCP clients that support stdio transport, like Claude Desktop.

To use stdio mode, simply run the server without the --http flag:

```bash
npx -y contentful-mcp --management-token YOUR_TOKEN