MCP for AI assistants

Content Library runs a Model Context Protocol server, so assistants such as Claude can search your catalog, explain why a product is not ready for a marketplace and fix product content — with the permissions of an API key you choose.

Connect

Setting	Value
Server URL	`https://library.retailcommerceai.com/api/mcp`
Transport	Streamable HTTP (stateless; JSON responses, no SSE stream or sessions)
Protocol version	`2025-06-18`
Authentication	`Authorization: Bearer cl_live_…`

Claude Code:

bash

claude mcp add --transport http content-library https://library.retailcommerceai.com/api/mcp \
  --header "Authorization: Bearer cl_live_YOUR_API_KEY"

Clients configured with JSON (headers supported):

json

{
  "mcpServers": {
    "content-library": {
      "type": "http",
      "url": "https://library.retailcommerceai.com/api/mcp",
      "headers": {
        "Authorization": "Bearer cl_live_YOUR_API_KEY"
      }
    }
  }
}

Choose the key's scopes deliberately

The assistant can do whatever the key allows. Use a read-only key (products:read, catalog:read, channels:read) for research and reporting; add products:write only when you want the assistant to edit products. Tools the key cannot use are not offered to the assistant.

Tools

Tool	Scope	What it does
`search_products`	`products:read`	Find products by text, status, category, collection, brand, tag or change date
`get_product`	`products:read`	Everything about one product by SKU; locale for translated content
`list_categories`	`catalog:read`	The category tree with codes, paths and product counts
`list_channels`	`channels:read`	Channels with ready, published, outdated and failing counts and top issues
`get_channel_payload`	`channels:read`	One product on one channel: mapped values, issues and the API payload
`update_product`	`products:write`	Change fields of a product (versioned, restorable)

Every tool has a JSON Schema for its arguments (returned by tools/list) and returns JSON as text and structured content.
update_product takes sku and changes — the same fields as PATCH /products/{sku} — plus an optional expectedVersion. Changes are recorded in the product history with the API key as the author.
Errors a model can fix (unknown SKU, invalid value) come back as tool results with isError, not as protocol errors.
Calls count against the key's rate limit like REST requests.

Raw JSON-RPC

Any HTTP client can talk to the server. Each POST carries one JSON-RPC 2.0 message:

bash

curl -X POST https://library.retailcommerceai.com/api/mcp \
  -H "Authorization: Bearer cl_live_YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -H "MCP-Protocol-Version: 2025-06-18" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_channel_payload","arguments":{"channel":"Amazon.ae","sku":"FR-OUD-100"}}}'

Supported methods: initialize, ping, tools/list and tools/call. Notifications (such as notifications/initialized) are accepted with 202. GET and DELETE answer 405 because the server keeps no sessions.