search_deep_agent
A search that reasons about the question, fans out over several phrasings and returns the pages with the passages that answer. ₹1.36 per call; a result that finds nothing is ₹0.
What it does
A deeper web search for a question that one query cannot answer: the source reasons about the question, fans out over several phrasings (yours plus additionalQueries) and returns data.results[] with url, title, author, publishedDate and highlights[], the passages that bear on the question, up to ten pages per call.
Use it for hard, multi-part questions; search_deep is the other thorough search (two or three keyword queries plus an objective, with Markdown excerpts), search_semantic a single search by meaning, answer_question one written answer with sources. Flat price per call for up to ten results (numResults 1 to 10, default 10); an empty result list is free. Three to eight seconds.
India: no country knob; put the place in the question.
Inputs
Generated from the contract captured from staging on 2026-11-05.
| Field | Type | Required | Description |
|---|---|---|---|
query | string | yes | The question, in words. |
numResults | integer | no | How many pages, 1 to 10. Default 10.; max 10 |
excludeDomains | array of string | no | Leave out given sites. |
includeDomains | array of string | no | Keep to given sites. |
endPublishedDate | string | no | Latest publication date, ISO 8601. |
additionalQueries | array of string | no | Up to five more phrasings the search fans out over. |
startPublishedDate | string | no | Earliest publication date, ISO 8601. |
What drives the cost
Flat per call for up to ten results; numResults above 10 is refused as invalid_input (free). A result that finds nothing is class: "ok_free", ₹0. The timeout is 60 seconds.
Price today
| Price today | ₹1.36 per call (cost_paise: 136) at ₹93.93 per US dollar on 17 November 2026 |
| Hold before the call | ₹2.04 (hold_paise: 204); when the call settles the real price is charged and the unused part is released |
| ₹0 when | it fails, or the provider answers but there is nothing in it (an empty list, no email, no page): class: "ok_free", ₹0 |
| Timeout | 60 s, then upstream_timeout (₹0, retryable) |
| Over MCP | run_tool with tool: "search_deep_agent" |
| Over REST | POST https://api.parel.si/v1/run with "tool": "search_deep_agent" |
| Confirmation | not needed (the estimate is under the ₹200 threshold) |
Request
curl https://api.parel.si/v1/run \
-H "Authorization: Bearer $PAREL_KEY" \
-H "Content-Type: application/json" \
-d '{"tool":"search_deep_agent","input":{"query":"how are Indian D2C fashion brands cutting their return rates","numResults":10}}'
What comes back
data, as the live check of 2026-11-05 recorded it, trimmed:
{
"results": [
{
"id": "https://www.firstresort.in/blogs/research/fashion-ecommerce-returns-rto-india-2026",
"title": "E-Commerce Returns & RTO in Indian Fashion 2026 | Data & Costs\n – First Resort by Ramola Bachchan",
"url": "https://www.firstresort.in/blogs/research/fashion-ecommerce-returns-rto-india-2026",
"publishedDate": "2026-08-15T12:43:16.000Z",
"author": "First Resort Editorial",
"highlights": [
"## 8. What measurably reduces returns\n\nThe fix list is short and quantified. Prepaid conversion: every order moved from cash to prepaid moves from a 20.9% to a 5.8% return probability — checkout incentives, UPI-first flows and COD verificat…"
]
},
{
"id": "https://shopflo.com/customers/manetain",
"title": "How Manetain reduced returns to 4% & shifted to 80% prepaid with Shopflo",
"url": "https://shopflo.com/customers/manetain",
"publishedDate": "2026-03-02T22:41:08.000Z",
"author": "Hinshara Habeeb, Co-founder, Manetain",
"highlights": [
"##### 1. COD was eating into profits\n\nAt the time, the payment split was simple:\n\n50% prepaid 50% COD\n\nReturns were around 17–18%.\n\n> “We were scaling on Meta… and without a limit on COD, it was going haywire.” — Hinshara Habeeb, Co-founder…"
]
},
{
"id": "https://shopflo.com/customers/past-modern-cs",
"title": "How a 2 year old apparel brand reduced returns by 67%",
"url": "https://shopflo.com/customers/past-modern-cs",
"publishedDate": "2026-08-08T04:37:06.000Z",
"author": "Shivang Chandna, Founder, Past Modern",
"highlights": [
"- Discounts and coupon offers were not clearly visible during checkout.\n\n...\n\n##### 1. Building bigger baskets without becoming a discount brand\n\nLike many fashion brands, Past Modern started with a standard first-purchase discount.\n\nBut as…"
]
}
]
}
Notes
- Flat per call for up to ten results; numResults above 10 is refused as invalid_input (free).
- Slower than search_semantic (three to eight seconds) because the source runs several searches and reasons over them; use it when one search is not enough.
- No country knob: name the place in the question or the extra phrasings.
Siblings
This page as Markdown: /docs/tools/search_deep_agent.md. Something wrong or missing: hello@parel.si.