Product Maestro over MCP Product Maestro v1.0.0

What this is

Connect

Claude Code, Cursor and other config-file clients

Add this to the client's MCP configuration file — .mcp.json for Claude Code — replacing the placeholder with the token you created in the app:

{
    "mcpServers": {
        "sappo-product-maestro": {
            "type": "http",
            "url": "https://api.sappo.xco.agency/product-maestro/mcp",
            "headers": {
                "Authorization": "Bearer mak_your_token_here"
            }
        }
    }
}

From the command line

claude mcp add --transport http sappo-product-maestro https://api.sappo.xco.agency/product-maestro/mcp --header "Authorization: Bearer mak_your_token_here"

Claude Desktop

Claude Desktop adds a remote server under Settings → Connectors → Add custom connector rather than from a configuration file, and it authenticates through the server's own flow instead of a header you type. This endpoint expects the Bearer token above, so use a client that lets you set the header: Claude Code, Cursor, or any client with a configuration file.

Any other MCP client

Point it at the endpoint above as a streamable-HTTP MCP server and have it send the Authorization header on every request. The store is identified from the token, so there is nothing else to configure.

What the assistant is told

The server sends these instructions to the client when it connects, so the assistant knows what it can do before it calls anything:

Product Maestro MCP server for accessing product schedule data from Shopify stores.

The store is identified automatically from your API token.

Available tools:
- `list_schedules`: List product schedules with optional status/archived filters and pagination
- `get_schedule`: Get a specific schedule by ID with its actions
- `list_actions`: List schedule actions with optional status/type filters and pagination
- `create_draft_schedule`: Create a draft schedule from structured actions (needs the schedules:write scope). Drafts never run on their own
- `execute_schedule`: Always refuses. Schedules are never run over the API

Drafting and running are deliberately separate. A draft runs only after the merchant reviews it in Product Maestro and clicks Save, so tell the merchant to do that rather than looking for a way to run it.

Tools

5 tools, read from the server itself. Every tool reads; none of them changes your data.

Try it

Once the server is connected, you do not call these tools yourself — you ask in words and the assistant picks the tool. Each row below is a question that reaches the tool beside it, so you can check the connection works and see the shape of an answer.

The assistant reads the store from your token, so none of these needs a shop name in it. If it answers without calling a tool, it is guessing: say “use the Product Maestro tools” and ask again.

When a tool fails

The MCP endpoint also has its own rate limit, separate from the REST API's per-token budget: 60 requests per 60 seconds per client address. Over it, the endpoint answers 429 with Too Many Attempts. — Laravel's own throttle response, not the API error envelope, so it carries no error.code — and the headers X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset, Retry-After.

An assistant that reports a tool "not found" is usually connected to the wrong app's endpoint. Compare the tool name it tried against the list above.