Connect an AI agent

What you are setting up

Fyre ships an MCP server that exposes scraping as a set of tools. Once your client knows about it, the agent can start extractions, page through results, query the resulting table, and export — all without raw HTML ever entering the conversation.

Claude Desktop

Open Settings, then Developer, then Edit Config, and add the server to the mcpServers block.

{
  "mcpServers": {
    "fyre": {
      "command": "npx",
      "args": ["-y", "@fyre/mcp"],
      "env": {
        "FYRE_LICENCE_KEY": "your-key-here"
      }
    }
  }
}

Claude Code and Cursor

One command, from anywhere:

claude mcp add fyre -- npx -y @fyre/mcp

What to ask for

Talk about the outcome, not the mechanics. The agent picks the tool.

• "Get every product on this category page with price and rating, follow pagination up to 20 pages."
• "Take the URL column from that table, visit each page, and add the spec fields."
• "How many of those rows are under £50 and in stock? Show me the top ten by rating."
• "Export the filtered set to CSV."

Why this stays cheap

Extraction results are held on Fyre's side as a table with a handle. The agent receives the schema and the row count — typically under 200 tokens — and then queries that table. Aggregations and filters run against the full data set but return only the answer. Pulling actual rows into context is an explicit step, and the agent is prompted to sample rather than dump.