strapi-mcp-setup - SKILL.md Agent Skill

name: strapi-mcp-setup description: 'Use when connecting an AI client to Strapi over MCP: the built-in content-management MCP server and the Kapa-powered Strapi Docs MCP server. Covers enabling, tokens, wiring.' tags: [strapi, mcp, ai, content-management, claude-code]

Strapi MCP Setup

Wire an AI client (Claude Code, Cursor, Claude Desktop) to Strapi's two MCP servers.

The Two Servers

Server	Purpose	Auth	Endpoint
strapi-mcp	Create/read/update/delete/publish content in your Strapi	Admin API token	`<your-strapi>/mcp`
strapi-docs	Query the Strapi documentation for accurate, current answers	None	`https://strapi-docs.mcp.kapa.ai`

The strapi cue profile wires both already (mcps: [strapi-docs, strapi-mcp]). This skill covers turning them on and authenticating.

Prerequisites

For strapi-mcp: a running Strapi v5.47.0+ instance and an Admin API token. The content tools available depend on that token's permissions.
For strapi-docs: nothing, it is hosted and unauthenticated.
Optional manual wiring uses npx mcp-remote.

strapi-mcp (content management)

1. Enable the server

Add the mcp block to your Strapi server config and restart:

export default ({ env }) => ({
  host: env('HOST', '0.0.0.0'),
  port: env.int('PORT', 1337),
  app: { keys: env.array('APP_KEYS') },
  mcp: { enabled: true },
});

The endpoint becomes available at /mcp (e.g. http://localhost:1337/mcp). It uses Streamable HTTP, is stateless, and only accepts POST (GET/DELETE return 405).

2. Create an Admin token

Admin panel → Settings → API Tokens → Create new token. Scope it to the content actions you want the AI to perform. Copy the value once.

3. Connect Claude Code

claude mcp add strapi-mcp --transport http http://localhost:1337/mcp \
  -H "Authorization: Bearer YOUR_ADMIN_TOKEN"

Restart Claude Code, then run /mcp to confirm strapi-mcp is connected.

Via the cue profile

The profile's registry entry reads two env vars. Set them (profile env: or your shell) before materializing:

export STRAPI_URL="http://localhost:1337"
export STRAPI_ADMIN_TOKEN="paste-admin-token-here"

The entry expands to an HTTP server at ${STRAPI_URL}/mcp with the bearer header. Permission changes (revoking/editing the token) take effect on the next request, there is no session to restart.

strapi-docs (documentation)

Zero-config. The cue profile wires it as an HTTP MCP at https://strapi-docs.mcp.kapa.ai. To add it elsewhere manually:

{ "mcpServers": { "strapi-docs": { "url": "https://strapi-docs.mcp.kapa.ai" } } }

Tip: prefix doc questions with Use the strapi-docs MCP server to answer: so the model queries live docs instead of stale training data.

Rules

Scope the Admin token to the minimum actions the AI needs; it gates every content tool.
Never commit the Admin token; pass it through env (STRAPI_ADMIN_TOKEN).
strapi-mcp needs the server running and v5.47+; until then it errors, strapi-docs still works standalone.
The content MCP is Beta; verify destructive actions (delete/unpublish) before letting an agent run them unattended.

Next Step

Once connected, use building-with-strapi for the data model the MCP tools operate on, or strapi-content-api to consume the same content over HTTP.