Storyblok and AI: Is there an MCP server for Storyblok?
Hosted by Storyblok, with OAuth sign-in. According to Storyblok, permissions are freely chosen and the server covers only the Management API.
Storyblok and AI at a glance
| Official MCP server | Yes, mcp.storyblok.com, Management API only, sign-in with OAuth or access token |
|---|---|
| Community MCP server | Yes, for example hypescale/storyblok-mcp-server (MIT), with a fixed EU endpoint |
| Interface | Management API (REST), Content Delivery API and GraphQL API, depending on the region of the space |
| AI in the product | Storyblok Agents in Storyblok Labs from Growth and other AI features |
| Vendor | Storyblok GmbH Headless CMS |
Status: November 2026. Checked against the vendor’s own sources, see Sources.
Storyblok describes itself, according to Storyblok, as a headless CMS for developers and marketers. If you want an AI assistant such as Claude to work with a space, you ask: is there an MCP server for it?
In short
Yes. Storyblok runs an official, centrally hosted MCP server. It covers the Management API, not the Content Delivery API. Sign-in uses OAuth, and the user releases permissions and spaces one by one. Deleting runs through a separate tool with confirmation. We found no plan restriction.
What Storyblok offers itself
The MCP server. Storyblok describes it in the documentation as a hosted, stateless HTTP endpoint at https://mcp.storyblok.com/mcp. Instead of one tool per endpoint, it offers generic tools: search and describe find and explain operations, execute_readonly runs GET, execute_mutating runs POST, PUT and PATCH, and execute_destructive runs DELETE. upload_asset and upload_asset_finish upload files. According to Storyblok, execute_destructive requires explicit confirmation from the user. An asset upload needs shell access, because the file goes to S3 with cURL. Storyblok recommends the server for one-off changes and prototypes. For repeatable bulk changes, it names the CLI.
Sign-in and rights. Storyblok recommends OAuth. On first use, the client opens a browser window. There the user signs in and chooses permissions and spaces. The OAuth metadata lists the authorization code flow with PKCE (S256), dynamic client registration and scopes per resource, such as stories:read, stories:write and stories:publish. Alternatively, the server accepts a personal access token as a bearer header. Storyblok advises a token limited to the spaces and rights needed.
Clients. The setup page names Claude Code, Claude, Cursor, VS Code, Codex and the Gemini CLI. According to Storyblok, Storyblok is listed in Claude’s connector directory. Alternatively, you can add the server as a custom connector. In organizations, access depends on the organization’s connector permissions, according to Storyblok: an owner enables Storyblok, and each member signs in individually. Read and write tools default to “Needs approval”. We could not read Claude’s directory page, because it blocks automated requests. So we do not know of any restriction by country or plan there. According to the setup page, any client with HTTP transport works. We found no statement from Storyblok about using the MCP server with ChatGPT. Storyblok’s page on Claude names stories, components, assets, releases and tags, among others.
Plans. The pricing page does not mention the MCP server, and the documentation names no plan requirement. According to the documentation, calls count against the Management API rate limit. The technical limits list 3 calls per second for Starter and 6 for Growth, Growth Plus, Premium and Elite.
APIs. The Management API is a REST API. It accepts a personal access token or an OAuth token. The Content Delivery API reads published content with an API access token. According to Storyblok, the GraphQL API is a read-only endpoint. The base URLs depend on the region of the space.
Data region. When you create a space, you choose the server location, according to Storyblok: Europe, United States, Canada or Australia. According to Storyblok, content delivery runs through a CDN regardless of this choice. For China, Storyblok offers separate infrastructure. On its security page, Storyblok names AWS data centers in North America (US and Canada), Europe (Germany) and Australia. The pricing page lists “Preferred Data Centers” only from Premium: in Premium the row is marked with a dollar sign, in Elite with “Yes”, and in Starter, Growth and Growth Plus with “Not available”. The page does not explain the feature, and the spaces documentation names no plan requirement for choosing the location. We found nothing on whether the region can be changed later. We also found nothing on whether the hosted MCP server treats spaces outside the EU differently.
Contracting party. The legal notice names Storyblok GmbH, Peter-Behrens-Platz 2, 4020 Linz, Austria, company register number FN 479743 f, as operator of the website. The terms do not give their contracting party by full company name. The self-service terms and the enterprise terms are governed by Austrian law. Enterprise customers in the Americas have separate terms with Delaware as venue, and Storyblok, Inc. in Wilmington is named there as DMCA contact. We did not check which company is the contracting party for customers in the Americas. Among the sub-processors, the data processing agreement names Amazon Web Services EMEA SARL for hosting and OpenAI Ireland Ltd for AI features, among others.
AI in the product. Storyblok Agents is an assistant inside the CMS. According to Storyblok, it sees and changes only what the user’s permissions allow, and every change needs the user’s approval. Users can build their own agents and schedule them. According to the press release, Storyblok Agents can be activated through Storyblok Labs on plans from Growth. Other AI features are AI translations, alt texts and SEO texts. Admins can set up their own AI provider with their own credentials, such as OpenAI, Google Gemini or Anthropic Claude. Use consumes AI credits, and according to the pricing page their amount depends on the plan.
What the community offers
Storyblok first had a locally run server. The repository storyblok/mcp-server is archived and points to the hosted server. Third parties offer free servers as well, for example hypescale/storyblok-mcp-server with 160 tools for the Management API (MIT, last commit March 2026). It runs locally with an access token and hard-codes the EU endpoint mapi.storyblok.com in its code. Spaces in other regions need a different base URL. Kiran1689/storyblok-mcp-server is also freely available (MIT, last commit June 2025). According to Composio, the commercial service offers its own Storyblok connection for agents. The page does not prove which rights it uses in Storyblok.
What this means for your company
- The official server can change and delete content. At sign-in, release only the spaces and rights the assistant needs.
- Test new workflows in an environment or a test space, not in the live space. Storyblok advises this too.
- If data should stay in the EU, choose Europe when you create the space. Ask Storyblok how the MCP server and the AI features process data.
- What reaches the AI assistant falls under its terms, not those of Storyblok.
MCP explained in detail, with the questions IT departments ask before they connect a server: Guide "What is MCP?"
What an MCP server is and how we build one: MCP servers by dectria
FAQ
Questions about Storyblok and AI
Is there an MCP server for Storyblok?
Can the assistant delete in Storyblok?
How does the assistant sign in?
Does the MCP server apply to all plans?
Can I choose a region at Storyblok?
Which company is behind Storyblok?
Storyblok has an MCP server.
We help you roll it out safely.
Which assistants are allowed, which rights they get, what they may change, and how this fits your data protection.
Discuss the rollout