1. Create a key

Request access — once approved, sign in and issue a key from your dashboard. The secret is shown exactly once — copy it before closing the panel.

What you should see: a key row in your dashboard, and the plaintext secret on your clipboard.

2. Connect — pick your protocol

The same key works on both protocols. Agent stacks connect the MCP server; backends and scripts call the REST API directly — replace YOUR_BUYARY_KEY with your key:

claude_desktop_config.json
{
  "mcpServers": {
    "buyary": {
      "type": "http",
      "url": "https://services.buyary.com/mcp/",
      "headers": {
        "Authorization": "Bearer YOUR_BUYARY_KEY"
      }
    }
  }
}
terminal
curl -s "https://services.buyary.com/v1/product-intelligence?product_name=iPhone%2017%20Pro%20Max" \
  -H "Authorization: Bearer YOUR_BUYARY_KEY"

What you should see: MCP — your client lists buyary as a connected server offering the tools. REST — the curl returns the intelligence JSON directly.

3. Make the first call

Ask your agent about a product — it calls the tool with the product name:

your agent, behind the scenes
get_product_intelligence(product_name: "iPhone 17 Pro Max")

What you should see: a synthesized answer with a Buyary Score, highlights and considerations, and per-source consensus — not a pile of review links.

Querying something obscure? When nothing confidently matches the catalog, the tool returns found: false with did-you-mean candidates instead of guessing — see the reference.