Search

POST /search

Runs a metasearch across many engines, then (on advanced depth) fetches and extracts clean markdown from each result. Returns clean, structured JSON.

Request

curl

curl -X POST https://api.crisps.ai/search \
  -H "Authorization: Bearer $CRISPS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "context rot long-context LLM degradation",
    "search_depth": "advanced",
    "max_results": 5,
    "topic": "general",
    "include_answer": false,
    "include_domains": ["arxiv.org"]
  }'

Python

import os, httpx

r = httpx.post(
    "https://api.crisps.ai/search",
    headers={"Authorization": f"Bearer {os.environ['CRISPS_API_KEY']}"},
    json={
        "query": "context rot long-context LLM degradation",
        "search_depth": "advanced",
        "max_results": 5,
        "topic": "general",
    },
    timeout=60,
)
print(r.json()["results"])

Node.js

const res = await fetch("https://api.crisps.ai/search", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.CRISPS_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    query: "context rot long-context LLM degradation",
    search_depth: "advanced",
    max_results: 5,
  }),
});
console.log((await res.json()).results);

Try it

POST /search Try it

API base URL API key

Request body { "query": "context rot long-context LLM degradation", "search_depth": "advanced", "max_results": 5 }

Send request

Body parameters

Parameter	Type	Default	Description
query required	string	—	The search query.
search_depth	string	advanced	basic (snippets only, 1 credit) or advanced (fetch + extract raw_content, 2 credits).
max_results	integer	10	Number of results to return.
include_raw_content	bool | string	true	true / "markdown" / "text" — include extracted page body. advanced only.
include_answer	bool	false	Return an LLM answer with [n] citations. Adds a credit surcharge.
chunks_per_source	integer	3	1–3. On advanced, return the top-N most relevant chunks per source.
topic	string	—	general, news, or finance.
time_range	string	—	day | week | month | year (or d/w/m/y).
days	integer	—	Look-back window for news.
start_date / end_date	string	—	YYYY-MM-DD bounds.
include_domains	string[]	—	Restrict results to these domains.
exclude_domains	string[]	—	Drop results from these domains.
country	string	—	Locale hint for the general topic.
exact_match	bool	false	Only quoted-phrase matches.
safe_search	bool	false	Filter explicit results.
include_images	bool	false	Include related image URLs.
include_image_descriptions	bool	false	Add short descriptions to images.
include_favicon	bool	false	Include each result’s favicon URL.
include_usage	bool	false	Add a usage: { credits } object to the response body.

Response

{
  "query": "context rot long-context LLM degradation",
  "results": [
    {
      "title": "Lost in the Middle: How Language Models Use Long Contexts",
      "url": "https://arxiv.org/abs/2307.03172",
      "content": "We analyze how LMs use long input contexts …",
      "score": 0.97,
      "raw_content": "# Lost in the Middle\n\n## Abstract\n…"
    }
  ],
  "response_time": 2.41
}

Field	Type	Description
results[].title	string	Result title.
results[].url	string	Canonical URL.
results[].content	string	Search snippet.
results[].score	number	0–1 relevance (cross-encoder reranked on advanced).
results[].raw_content	string?	Extracted main content (markdown/text). Present on advanced when ≥ 200 chars.
answer	string?	Present when include_answer is set and a model is configured.
images	string[]?	Present when include_images is set.
usage	object?	{ credits } — present when include_usage is set.
response_time	number	Server processing time (seconds).

Tip

Credits are reported in the response headers — X-Credits-Used, X-Credits-Remaining, and X-Credits-Limit — not in the body. advanced searches are cached per query and per URL, so repeated queries return near-instantly.