Google web search in one call

Organic results, answer boxes, knowledge graphs and "people also ask" from Google — plus a Markdown output mode tuned for LLM context windows. Just say num=N (up to 200). 1 token per 10 results requested. Google delivers up to 10 organic per page — sometimes fewer when Answer Box, News, Images etc. take slots.

num

cost: 1 token

🔍 Try our API: curl 'https://api.lynk.run/search/v1/web?q=claude+code&format=markdown' click to copy

⚡

Fast & predictable

Backed by serper.dev — typical response in 1–2 s, and identical repeat queries come back near-instantly thanks to an internal cache layer.

🤖

LLM-friendly Markdown

Add format=markdown and the API returns a rendered document instead of JSON — perfect for feeding directly into an LLM context window without a parsing step.

🌍

Geo + language

Set gl (country) and hl (UI language) to localise results. Defaults to de/de — switch to us/en or any other Google-supported combination on a per-request basis.

API Reference

Tell us how many results you want with num (1–200). We paginate Google internally and stitch the results. Cost: 1 token per 10 results requested (num=10 → 1 token, num=50 → 5 tokens, num=200 → 20 tokens). Unauthenticated: 20 / minute / IP, no token cost.

GET — single query

curl 'https://api.lynk.run/search/v1/web?q=claude+code+best+practices&num=30' \
  -H 'Authorization: Bearer YOUR_API_KEY'

POST — batch (max 10 queries)

curl https://api.lynk.run/search/v1/web \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "queries": ["lynk run", "serper.dev", "claude opus 4.7"],
    "gl": "de",
    "hl": "de",
    "num": 30
  }'

Markdown output (LLM-friendly)

Pass format=markdown (query string or JSON body) and the response is rendered as a Markdown document instead of JSON. Drop it straight into a prompt — no parsing needed.

curl 'https://api.lynk.run/search/v1/web?q=lynk+run&format=markdown' \
  -H 'Authorization: Bearer YOUR_API_KEY'

Response shape (JSON)

{
  "query": "lynk run",
  "gl": "de",
  "hl": "de",
  "num": 28,              // actually returned (may be lower than requested)
  "requested": 30,        // what you asked for
  "fetched_at": 1779480000000,
  "organic": [
    { "title": "Lynk", "link": "https://lynk.run", "snippet": "…", "position": 1 }
  ],
  "answer_box": { "title": "…", "answer": "…", "link": "…" },
  "knowledge_graph": { "title": "…", "description": "…", "website": "…" },
  "people_also_ask": [{ "question": "…", "snippet": "…" }],
  "related_searches": ["…"]
}

Note: answer_box, knowledge_graph, people_also_ask and related_searches are only present when Google returns them. Absent ≠ empty — your code shouldn't assume they exist.

🤖 Use with AI / LLM Agents

Web Search is exposed as the web_search MCP tool. Connect any MCP-capable agent (Claude Desktop, Cursor, Cline) — the full spec lives at:

https://api.lynk.run/mcp/llms.txt