Contentful MCP Server

The Contentful MCP Server connects the headless CMS Contentful to AI assistants such as Claude Code, Codex, or Cursor. Instead of clicking from entry to entry in the web interface, you describe in natural language what should happen, and the assistant runs the matching calls against the Contentful Management API. The server is developed and maintained by Contentful itself as an open-source project under the MIT licence; the repository was active and not archived on the day of review.

What is the Contentful MCP Server?

Contentful is a headless CMS: content exists as structured entries whose shape is defined by content types, and it is delivered to any frontend via API. The MCP server opens exactly this management layer to language models. It is started locally via npx as a Node.js process and talks to the Management API of your space. The assistant therefore works with the real content and models rather than with an export.

What can the server do?

According to the README, the toolset comprises well over forty functions across several areas. For content types there are create, update, publish, and delete. For entries you can search, edit, publish, unpublish, and delete, also in bulk, plus a semantic search by meaning and retrieval of version snapshots. Assets can be uploaded, described, and published. Further areas are spaces and environments, locales for multilingual content, tags, content preview URLs, and releases including validation and scheduled publishing. In addition there are AI Actions, Contentful-provided AI workflows the assistant can create and invoke, and a query for the usage metrics of an organisation.

Typical use cases

The server is particularly helpful for tasks that mean many uniform clicks in the web interface: adding a field to a content type and adjusting existing entries, drafting an entry for a product launch, uploading images and organising them by campaign, or preparing content for a new locale. Development teams can sketch and review content models from the chat before using them in code. Editorial teams benefit from bundling publications as a release and having it validated in advance.

Setup and prerequisites

You need Node.js and npm, a Contentful space ID, and a personal access token for the Management API. The server is registered as the command npx -y @contentful/mcp-server; you pass token, space ID, environment, and host as environment variables. There is a one-click link for Cursor, a .dxt file in the GitHub releases for Claude Desktop, and a ready-made codex mcp add command for Codex.

Security and protected environments

Because the server can write and delete, the PROTECTED_ENVIRONMENTS option matters: when set, the server blocks write and delete calls in the listed environments, such as master, before they reach the API. According to the README, the protection applies only inside this server and does not replace role-based permissions. Work in a separate test environment where possible and use a token with narrow rights.

Costs and licence

The MCP server itself is free and open source under the MIT licence. A Contentful account is required, however; which features are included in which plan, for example AI Actions, is governed by Contentful's official pricing page. Please check there whether and to what extent a free tier exists.

Who is the Contentful MCP Server for?

It suits teams that already use Contentful and want to speed up recurring maintenance, migrations, or modelling tasks. It offers no benefit to beginners without a Contentful account. Anyone using it should always review changes to production content, because a language model can fill fields incorrectly or change too many entries at once.

Frequently asked questions

Can the server delete content by accident? Yes, it has delete tools. Protect important environments via PROTECTED_ENVIRONMENTS and use a token with limited rights.

Do I need programming skills? Basic knowledge of configuring an MCP client is enough; the actual work happens in natural language.

Is this an official product? Yes, the repository lives in Contentful's GitHub organisation.

Node.js and npm, a Contentful account with a space ID and a personal Management API access token, plus an MCP client such as Claude Code, Codex, Cursor, or Claude Desktop.

With the Contentful MCP Server, an AI assistant manages content and content models of your space. This guide leads from the prerequisites to the first safe request.

Step 1: Check the prerequisites

You need Node.js with npm, a Contentful account, the space ID, and an MCP client such as Claude Code, Codex, or Cursor. You can find the space ID in the settings of your space.

Step 2: Create an access token

Create a personal access token for the Management API in Contentful. It inherits your rights, so use it only for spaces in which the assistant is allowed to work. Keep the token only in your client configuration.

Step 3: Register the server

Register the server as the command npx -y @contentful/mcp-server and set the environment variables CONTENTFUL_MANAGEMENT_ACCESS_TOKEN and SPACE_ID. With ENVIRONMENT_ID you define which environment is used; the default is master. For Codex, the README gives a ready-made codex mcp add command; for Cursor, a one-click link.

npx -y @contentful/mcp-server
codex mcp add

Protect production environments

Set PROTECTED_ENVIRONMENTS to a comma-separated list such as master,staging. Write and delete calls in these environments are then blocked, while reads remain possible. Note that this protection applies only inside the server.

Step 4: Test the connection

Start with a read-only request, such as "List all content types". Then have a draft created in a test environment and check it in the web interface.

Working day to day

Ask the assistant to first list which entries would be affected before larger changes. Bundle publications as a release and have it validated in advance.

Personal Contentful Management API access token (CONTENTFUL_MANAGEMENT_ACCESS_TOKEN) as an environment variable.

Read and write access to content types, entries, assets, locales, tags, releases, environments, and AI Actions in the configured space, limited by the rights of the token used.

The server runs locally and calls the Contentful API. Content the assistant retrieves enters the AI client context and therefore reaches the model provider.

Accidental overwriting, publishing, or deleting of content and models, overly broad token rights, and deletion of entire environments; protection only inside this server via PROTECTED_ENVIRONMENTS.