Image search for AI agents: direct URLs, thumbnails, and visual context

Web search tells an agent what's been said. Image search tells it what something looks like. /images returns direct image URLs, thumbnail previews, and the source page — content a multimodal agent can fetch, display, or reason over.

One call, images out

curl "https://superhighway.walls.sh/images?q=golden+gate+bridge&limit=5"

The response:

{
  "query": "golden gate bridge",
  "count": 5,
  "results": [
    {
      "title": "Golden Gate Bridge at sunset",
      "url": "https://example.com/page-with-image",
      "img_src": "https://example.com/images/golden-gate.jpg",
      "thumbnail": "https://tbn.example.com/th?q=golden+gate",
      "format": "jpeg",
      "engines": ["bing images"]
    },
    ...
  ]
}

Each result has: img_src — the direct image URL the agent can fetch or pass to a vision model; url — the source page for attribution; thumbnail — a smaller preview (may be null); format — the image format (jpeg, png, webp, …). Price is $0.001 per call.

What agents use it for

• Visual grounding — pair image_search with a vision model to look up what a concept actually looks like, not just read about it.
• Content generation — an agent writing a blog post or presentation finds reference images to describe, embed, or license.
• Multimodal research — combine with web_search or /research to answer both "what does X mean?" and "what does X look like?"
• Product and brand monitoring — search for a brand name to surface visual mentions across the web.

Use it via x402 (no signup)

An agent with a funded Base wallet and an x402 client pays per call automatically — no API key, no account:

import { wrapFetchWithPayment, createSigner } from "x402-fetch";
const signer = await createSigner("base", process.env.AGENT_PRIVATE_KEY);
const pay = wrapFetchWithPayment(fetch, signer);

const res = await pay("https://superhighway.walls.sh/images?q=space+nebula&limit=5");
const { results } = await res.json();
console.log(results[0].img_src); // direct image URL

Full flow: A web search API your AI agent can pay for by itself.

Use it via MCP (one command)

If your agent speaks MCP (Claude Code, Cursor, Windsurf), install the server once and call the image_search tool — payment is handled internally:

claude mcp add superhighway -e AGENT_PRIVATE_KEY=0x... -- npx -y superhighway-mcp

Then: "Search for images of the Eiffel Tower at night." — no HTTP, no wallet code, just the tool.

Parameters

Full API reference: /openapi.json. All five tools available in the MCP server.