What is the Contentful MCP server?
The Contentful MCP server is an MIT licensed Node.js server that connects an MCP client to the Contentful Management API. It offers tools for content types, entries, assets, spaces, environments, locales, releases, tags and AI actions, and can block writes to named environments.
It is a Model Context Protocol server for the Contentful Management API, published in the contentful organisation on GitHub. An agent connected to it can create and edit content in a space using plain language requests.
The README gives example asks such as creating a blog post for a product launch, updating every product entry with new pricing, adding a field to a content type, or translating content with an AI action.
What an agent can do with it
- Search, read, create, patch, publish and delete entries, with version history snapshots.
- Upload and manage assets and their publish state.
- Change content types, locales, tags, spaces and environments.
- Build releases and schedule them to publish or unpublish later.
- Create and invoke AI actions inside a space.
Who it is for
Best for
- Content teams who run repetitive edits across many entries.
- Developers prototyping a content model by asking an agent to create types and fields.
- Editors who need to translate or localise content in bulk.
Not for
- Read only delivery of published content to a website. This server targets the Management API.
- Teams without a personal access token and a space ID to give it.
Use it with UiChemy
Contentful stores the content. When the front end is a WordPress site, UiChemy's Figma to WordPress converts a design into an editable page, and Composer writes HTML, CSS and JS. Click Get Started on uichemy.com.
Connect Contentful MCP Server: Content Tools for AI Agents to your AI client
npx -y @contentful/mcp-serverCollect the prerequisites
You need Node.js and npm, a Contentful account with a Space ID, and a Content Management API personal access token.
Pick an install route
Use the Cursor one click link, import the dxt file from the latest release into Claude Desktop, run the codex mcp add command, or paste the config above.
Protect important environments
Set PROTECTED_ENVIRONMENTS to a comma separated list such as master,staging. Write and delete calls to those environments are then refused.
Run a read first
Ask for a list of content types before any write, and rehearse changes in a non production environment.
Client configuration
{
"mcpServers": {
"contentful-mcp": {
"command": "npx",
"args": ["-y", "@contentful/mcp-server"],
"env": {
"CONTENTFUL_MANAGEMENT_ACCESS_TOKEN": "your-CMA-token",
"SPACE_ID": "your-space-id",
"ENVIRONMENT_ID": "master",
"CONTENTFUL_HOST": "api.contentful.com"
}
}
}
}Endpoints and ports
your-CMA-tokenRequired. Management API personal access token.your-space-idRequired. The space to work in.masterOptional. Defaults to master.master,stagingOptional. Blocks writes and deletes there.api.contentful.comOptional. Defaults to api.contentful.com.What you can do with it
- Protected environmentsAn opt in guard blocks write and delete tools on environments you name, while reads keep working.
- Semantic searchFind entries by meaning, not only by field filters.
- Entry snapshotsRetrieve version history for an entry before you change it.
- Releases and schedulingGroup entries and assets into a release and schedule publishing or unpublishing.
- Locale managementCreate and update locales so translation work can be driven from the agent.
- AI actionsCreate, publish and invoke AI actions that run inside Contentful.
- Several install routesOne click Cursor link, a desktop client dxt file, a Codex command or a plain npx config.
Tools it exposes
get_initial_context
Starts the connection and returns usage instructions.
list_content_types / create_content_type / update_content_type
Read and change content models.
search_entries / semantic_search
Filter entries or find them by meaning with vector search.
create_entry / update_entry / patch_entry
Write entries, or change specific field locales with JSON Patch.
publish_entry / unpublish_entry / delete_entry
Control entry state, in bulk where supported.
upload_asset / list_assets / publish_asset
Handle media files and their metadata.
list_locales / create_locale
Manage languages for multi language content.
create_release / publish_release / schedule_publish_release
Group changes and publish them now or later.
create_ai_action / invoke_ai_action
Build and run AI actions in a space.
get_usages
Returns aggregated usage metrics for an organisation.
Frequently asked questions
What is the Contentful MCP server?
It is a Model Context Protocol server that exposes the Contentful Management API as tools an AI client can call. It covers content types, entries, assets, spaces, environments, locales, releases, tags and AI actions, and it is released under the MIT licence.
How do I install the Contentful MCP server?
Run it with npx and the package @contentful/mcp-server, passing a Management API token and your space ID as environment variables. The README also offers a Cursor link, a dxt file for Claude Desktop and a Codex command, or you can build from source.
What credentials does it need?
A Contentful Management API personal access token and a Space ID are required. The environment defaults to master and the host to api.contentful.com. Treat the token like a password, because it can change content and delete entries across your spaces.
Can I stop the agent from writing to production?
Yes. Set PROTECTED_ENVIRONMENTS to the environment IDs you want locked, for example master. Writes and deletes there return an error. The README notes the guard only runs inside this server, so it does not replace role based permissions.
Does it publish content automatically?
It can publish entries, assets and releases when asked, and it can schedule a release for a later time. It acts on your instructions, so word requests carefully and review drafts before telling the agent to publish.
Can it read the Content Delivery API?
The README names the Management API only. For a front end that just reads published content you would use the delivery or preview APIs directly. This server suits editing, modelling and release work.
Is it free to use?
The server is MIT licensed, so the software is free. Contentful plans and limits are a separate matter and the README gives no pricing. Check your plan before bulk operations, because large imports and AI actions may count against usage.
What should I read next?
If your site is WordPress, see WP MCP Ultimate for content tools, Respira for page builders, and the WooCommerce MCP server for store data. Each covers a different part of a site, so pick by the job you need done.


