What is the Strapi MCP server?
The Strapi MCP server is a community Node.js server that connects an MCP client to one or more Strapi instances. It reads content types and components, runs REST requests, uploads optimised media and applies a write protection policy. The README flags it as AI assisted and untested for production.
This is a community built Model Context Protocol server for Strapi CMS. It sits between an AI client and your Strapi instance, reading the schema first and then calling the REST API on your behalf.
Instead of many narrow tools it uses a few broad ones. A single REST tool takes an endpoint, method, parameters and body, which means any collection your API token can reach is available to the agent.
What an agent can do with it
- List configured Strapi servers and read content types and components, with paging.
- Query, create, update and delete content through one REST tool, with filters, sorting and populate.
- Upload media from a URL, converted to a chosen format and quality.
- Work with several Strapi instances from one config file.
- Get version aware hints for Strapi v4 and v5.
Who it is for
Best for
- Developers exploring a Strapi schema by asking questions about it.
- Teams that run several Strapi environments and want one connector.
- Editors prototyping bulk content edits on a development instance.
Not for
- Production sites without testing. The README says not to use it in production without thorough validation.
- Anyone who needs a hosted endpoint. It runs locally from a config file.
Use it with UiChemy
Strapi serves content to whatever front end you build. If that front end is a WordPress site, UiChemy's AI to WordPress imports AI builds, and Composer writes the HTML, CSS and JS. Click Get Started on uichemy.com.
Connect Strapi MCP Server: REST Access to Your Strapi CMS to your AI client
npx -y @bschauer/[email protected]Register the server in your client
Add an mcpServers entry that runs npx with the package @bschauer/[email protected] as shown in the README.
Create an API token
Log in to the Strapi admin panel and create an API token with only the permissions you need.
Write the config file
Save the block above at ~/.mcp/strapi-mcp-server.config.json. Add more named entries for more Strapi instances, and set a version such as 5.* if you like.
Inspect before writing
Ask the agent to read content types first. Use a development instance for any create, update or delete request.
Client configuration
{
"myserver": {
"api_url": "http://localhost:1337",
"api_key": "your-jwt-token-from-strapi-admin",
"version": "5.*"
}
}Endpoints and ports
http://localhost:1337Base address of the Strapi instance.your-jwt-token-from-strapi-adminToken from the Strapi admin panel.5.*Optional. Examples given are 5.*, 4.1.5 and v4.~/.mcp/strapi-mcp-server.config.jsonWhere the server reads its settings.What you can do with it
- Schema introspectionThe agent can read content types and components before it writes anything.
- Single REST toolOne call covers GET, POST, PUT and DELETE, with filters, sort, pagination and populate.
- Write protection policyCreate, update, delete and media upload are described as protected operations that need explicit authorisation and are logged.
- Media handlingImages can be fetched from a URL, converted and compressed before upload.
- Multiple serversSeveral Strapi instances live in one config file under their own names.
- Version awarenessSettings can name a Strapi version, and the server adapts guidance for numeric IDs in v4 and document IDs in v5.
Tools it exposes
strapi_list_servers
Lists configured servers with version information.
strapi_get_content_types
Returns content types from a chosen server.
strapi_get_components
Returns components, with page and pageSize options.
strapi_rest
Runs a REST call with endpoint, method, params and body, checked against the schema.
strapi_upload_media
Uploads an image from a URL with format, quality and metadata such as alt text.
Frequently asked questions
What is the Strapi MCP server?
It is a community Model Context Protocol server that lets an AI client work with a Strapi CMS. It exposes schema reading, a general REST tool and media upload, and it is released under the MIT licence. It is not a Strapi product, so check its issues before depending on it.
How do I install the Strapi MCP server?
Add an mcpServers entry that runs npx with the package @bschauer/strapi-mcp-server pinned to version 2.6.0, then create a config file at ~/.mcp/strapi-mcp-server.config.json with your API address and token. Restart the client to load the five tools.
Is the Strapi MCP server safe for production?
The README warns that the software was built with AI assistance, is provided as is and should not be used in production without thorough testing and validation. Use it on development instances first and keep a database backup before any write.
Does it work with Strapi v4 and v5?
The README says the config can name a version and describes handling differences, such as numeric IDs and nested attributes in v4 against document based IDs and flat data in v5. Test your own collections, since behaviour can differ by plugin.
What credentials does it need?
It needs the Strapi API address and a token created in the admin panel, stored in a local config file. The README calls it a JWT token. Limit the token's permissions to the collections you want the agent to touch.
Can the agent upload images to Strapi?
Yes. The media tool takes an image URL, an output format such as webp, a quality value and metadata including name, caption and alternative text. Uploads count as protected write operations under the README's policy.
Can I connect more than one Strapi site?
Yes. The config file holds named entries, one per instance, and each tool call names the server it targets. A tool lists the configured servers together with their version information.
What other CMS servers are there?
For WordPress content, see WP MCP Ultimate. For store data there is the WooCommerce MCP server, and Lara Dashboard MCP covers a Laravel CMS. Each one targets a different system, so choose by the platform you actually run.


