Pixserp

pixserp is an AI-native search MCP server. Add it to your client and your AI assistant gets access to the live web — with structured citations, across ten answer shapes, from a single tool call.

What you get

• One tool, ten answer shapes — web, news, images, places & maps, shopping, flights, hotels, YouTube, transcripts, any URL. The right shape is picked automatically from your query.
• Cited answers from the live web — every response comes back with inline [1] markers and a structured citations array (URLs, titles, snippets, plus per-shape fields like rating, price, hours, GPS).
• Four research depths — pixserp-fast for quick lookups, pixserp-standard for general queries, pixserp-deep for multi-angle research, pixserp-agent for multi-step research loops that decide when to stop.
• Drop-in for the rest of your stack — same backend powers an OpenAI-compatible REST API (/v1/chat/completions, /v1/responses), so you can mix MCP usage in your IDE with API usage in your code.
• Flat per-request pricing, predictable cost, $2.50 free welcome balance, no card to start.

Setup

Paste the config below, replace pxs_YOUR_KEY with your key from pixserp.com/api-keys, restart your client. The server appears as a tool named search.

{
  "mcpServers": {
    "pixserp": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://pixserp.com/api/v1/mcp",
        "--header", "Authorization: Bearer pxs_YOUR_KEY"
      ]
    }
  }
}