SearchPipe is a search API built for AI agents and applications: one POST /search call runs multi-engine retrieval, full-text page extraction, LLM reranking and optional summarization — returning structured, scored results your code can consume directly. A Tavily alternative with Chinese web coverage.
A search API lets your application or AI agent query the web programmatically and receive structured results — titles, URLs, snippets, scores — instead of raw HTML. Modern AI-powered search APIs go further: they fetch the actual page content, filter out navigation and ads, rerank results by relevance using large language models, and optionally generate cited answers.
SearchPipe is exactly that: an AI search API designed for agents. You send a query; we run the entire pipeline — multi-engine retrieval via SearXNG, full-text extraction, LLM reranking (0–1 score), optional AI answer with citation markers — and return clean JSON your agent can act on without any parsing.
Building a production search pipeline means gluing together search engines, crawlers, extractors, rerankers and moderators — then maintaining them forever. A search API removes that burden:
Aggregates multiple search engines — self-hosted SearXNG, fully under our control, no third-party search API fees.
Automatically fetches top-N pages and extracts the article body, filtering out navigation, ads and noise.
A large model scores relevance (0–1) against the query intent, so the most relevant results come first.
Optionally generates a direct answer with citation markers, giving your agent ready-to-consume factual context.
Every call is authenticated with an sp- prefixed API key (create one). Two ways to pass it:
Authorization: Bearer sp-YOUR_KEY
# or
X-API-Key: sp-YOUR_KEYcurl -X POST https://searchpipe.tech/search \
-H "Authorization: Bearer sp-YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"query":"FastAPI tutorial","max_results":5,"include_answer":true}'import httpx
resp = httpx.post(
"https://searchpipe.tech/search",
headers={"Authorization": "Bearer sp-YOUR_KEY"},
json={"query": "FastAPI tutorial", "max_results": 5, "include_answer": True},
timeout=60,
)
data = resp.json()
print(data["answer"])
for r in data["results"]:
print(r["score"], r["title"], r["url"])const resp = await fetch("https://searchpipe.tech/search", {
method: "POST",
headers: {
"Authorization": "Bearer sp-YOUR_KEY",
"Content-Type": "application/json",
},
body: JSON.stringify({ query: "FastAPI tutorial", max_results: 5, include_answer: true }),
});
const data = await resp.json();| Parameter | Type | Default | Description |
|---|---|---|---|
query required | string | — | Search question or keywords |
max_results | int | 5 | Number of results (1–20) |
include_answer | bool | false | Generate a cited AI answer |
include_raw_content | bool | false | Return the full extracted page body |
search_depth | string | basic | Search depth (basic / advanced); affects credit cost |
The response is structured JSON; every result carries a 0–1 relevance score and results are sorted by relevance, highest first:
{
"query": "How to deploy FastAPI",
"answer": "FastAPI is typically served with uvicorn,
often behind gunicorn…[0][2][3]",
"ai_generated": true,
"results": [
{
"url": "https://…",
"title": "…",
"content": "Body snippet…",
"score": 0.98,
"raw_content": "Full page body (only when include_raw_content=true)"
}
]
}Give Claude Code, Cursor or your custom agent real-time web search with one API call — no MCP setup required, just POST /search.
Run scheduled queries and land scored article bodies in your own store for digests, sentiment and price tracking — cached hits keep costs predictable.
Use include_answer to get a cited summary and turn "I can't answer time-sensitive questions" into sourced answers.
Batch queries in parallel with automatic extraction — no crawler or anti-bot maintenance on your side; failures are auto-refunded, retries are safe.
Credit-based, pay-as-you-go: $0.005 per credit (limited-time 50% OFF; standard $0.01). Basic search costs 1 credit per call, advanced search 2 credits. New accounts get 1,000 free credits per month. Recharged credits never expire; subscription credits are valid for 30 days. Full details on the Pricing page.
SearchPipe is an AI search API designed for agents and applications that need structured, scored, cleaned results — not raw search-result HTML. It goes beyond traditional search APIs by adding full-text extraction, LLM reranking and optional cited answers.
Same developer-experience style (API key, one endpoint, structured JSON), but SearchPipe adds Chinese web coverage, per-call credit pricing with automatic refunds, and a remote MCP Server option for zero-code agent integration.
Yes. The REST API (POST /search) is fully independent — use it from any language, framework or backend service. MCP is just a convenience layer for agent clients.
Nothing. Upstream retrieval failures (502) and output content violations (400) are automatically refunded (idempotently). Only successful requests are billed.
Default: 100 requests per minute with a burst of 20, enforced by a Redis sliding window. Over-limit requests return 429 with a Retry-After header. Contact us to raise the quota for high-volume use.
Free credits on signup — make your first API call in under a minute.