Storyblok MCP Server

The Storyblok MCP Server connects your AI assistant directly to a space in the Storyblok headless CMS. MCP, the Model Context Protocol, is an open standard that lets assistants such as Claude, Cursor or Codex work with external services. A headless CMS manages content in one place and delivers it to websites and apps through an API. The server comes from Storyblok itself, a company founded in Austria in 2017 and registered in Linz. Storyblok hosts the server at mcp.storyblok.com. There is nothing to install: you add an address to your client and sign in.

What is the Storyblok MCP Server good for?

Everyday content work is full of small jobs that take a lot of clicks in the backend. Think of finding every story in a folder that is missing a certain block. Or adding a new field to a component, cleaning up tags and uploading an image. With the MCP server, you describe tasks like these in plain language. The assistant looks up the right Management API operation and runs it. The Management API is the interface for managing a space's content, component schemas, assets and settings. According to Storyblok, the server is best suited to one-off changes and exploratory work: looking things up, checking them and making targeted edits.

How the server works

Instead of exposing hundreds of separate tools, the server works in three steps. First, the search tool finds the right operation by keyword. Next, describe returns that operation's parameters and schema. Finally, the assistant executes it. There are three distinct tools for that: one for read-only access, one for creating and updating, and one for deleting. According to the documentation, deletions require explicit confirmation from the user. This split is handy because it lets you decide in your client which kind of access may run without asking. A fields filter keeps large responses small. Assets have their own upload tool. The final upload step does need shell access, though, because the file is sent with a cURL command.

Limits worth knowing

Storyblok is refreshingly candid in its documentation. The server only talks to the Management API. It does not read published content through the Content Delivery API. Because a language model sits between you and every request, results are not deterministic. For bulk changes, CI/CD pipelines and migrations, Storyblok explicitly recommends its CLI instead. The usual Management API rate limits apply as well.

Sign-in and clients

OAuth is the recommended way to sign in. On first connect, your browser opens, you log in to Storyblok and choose exactly which permissions and spaces the assistant may use. For scripts or clients without OAuth support, a personal access token sent as a bearer header is the alternative. Storyblok provides dedicated instructions for Claude Code, Claude via its connector directory, Cursor, VS Code, Codex and the Gemini CLI. Any other client with HTTP transport uses the same address.

Cost and license

The documentation mentions no separate fee and no minimum plan for the MCP server. According to its pricing page, Storyblok itself offers a free Starter plan for testing and personal projects, alongside paid plans and enterprise tiers sold through its sales team. The hosted server has not been published as open source. An earlier GitHub repository from its experimental phase has since been archived.

Who is the Storyblok MCP Server for?

The server suits content teams and developers who already use Storyblok and want to get routine work in a space done faster. Developers can review or adjust component schemas without switching between the editor and the API reference. Editors find content more quickly and can have inconsistencies tracked down for them. Compared with the MCP servers of other CMSs such as Contentful, Sanity or Strapi, the clear split between read, write and delete access stands out. If you regularly move large amounts of data, stick with the CLI.

A Storyblok account with access to at least one space and a client that supports remote MCP over HTTP, such as Claude Code, Claude, Cursor, VS Code, Codex or the Gemini CLI. For scripts, a personal access token as an alternative.

With the Storyblok MCP Server, your AI assistant manages a Storyblok space's content, components and assets through the Management API. Storyblok hosts the server at https://mcp.storyblok.com/mcp. This guide covers setup in common clients and what to watch for with permissions.

Step 1: Sort out sign-in first

OAuth is recommended. On first connect, your browser opens, you log in to Storyblok and select permissions and spaces. For scripts or clients without OAuth, create a personal access token instead. Scope it to the spaces and rights you need and keep it in an environment variable, not in your repository.

Claude Code

Run claude mcp add --transport http Storyblok https://mcp.storyblok.com/mcp. Use --scope user or --scope project to decide where the entry applies. Then open /mcp, pick Storyblok and sign in in the browser.

Claude

Open Settings, Customize and Connectors, search for Storyblok via Browse connectors and click Connect. In organizations, an owner has to enable the connector first. If Storyblok is not in the directory, add a custom connector with the server address and leave the OAuth fields empty.

Cursor and VS Code

In Cursor, add an entry called Storyblok with the url https://mcp.storyblok.com/mcp under mcpServers in .cursor/mcp.json. In VS Code, the key in .vscode/mcp.json is servers rather than mcpServers, and the entry also needs type http. In both editors, confirm the sign-in afterwards.

Codex and Gemini CLI

For Codex, add an mcp_servers.storyblok section with the url to ~/.codex/config.toml and start sign-in with codex mcp login storyblok. In the Gemini CLI, add the address as httpUrl in ~/.gemini/settings.json and set authProviderType to dynamic_discovery.

Work safely

Grant only the rights and spaces the assistant really needs. Have planned operations described before they run, and check IDs and block names before deleting anything. Test new routines in a separate environment first. For bulk changes and migrations, the Storyblok CLI is the better tool.

claude mcp add --transport http Storyblok https://mcp.storyblok.com/mcp

OAuth in the browser (recommended), where you choose permissions and spaces. Alternatively, a personal access token sent as a bearer header, scoped as tightly as possible to the spaces and rights you need.

Read, create, update and delete via the Management API, within the permissions you grant: stories, component schemas, assets, tags and other space settings. Deletes only after explicit confirmation.

Requests run through the server hosted by Storyblok to the Management API. Space content shown in the chat also reaches the respective AI provider. According to Storyblok, tokens should never be committed to version control and must be revoked immediately if leaked.

Medium. The assistant can change or delete content and schemas. A crafted text with hidden instructions can trigger unwanted changes. Storyblok recommends minimal permissions, reviewing planned operations, checking IDs before deletes and testing in a separate environment first.