/research

Deep research for a cent a report

One call: a deep search, the topic's sub-topics and the pages its lists link to, returned as a cited report of quoted passages. $0.01 a report. No LLM, nothing made up.

Read the docs Try search first

report.markdown

# Research: How do vector databases index embeddings?

## Summary

> HNSW builds a layered graph in which each vector links to its nearest neighbours ... [1]

## Key facts

- Product quantization splits each vector into m sub-vectors ... [2]

## How it works

> At query time the search starts at the top layer and greedily ... [1]

## Sources

[1] Hierarchical navigable small world. en.wikipedia.org

Quickstart

One request, a whole report

depth deep on /v1/research, or Tavily's research endpoint with the Tavily SDK. The report comes as Markdown and as JSON (summary, keyFacts, sections, sources).

sources in numbered, APA, MLA or Chicago style
outputSchema fills your JSON fields from the same pages
async: true for a job id when your agent should not wait

Research reference

curl -s https://api.contextfetch.dev/v1/research \  -H "Authorization: Bearer $CONTEXTFETCH_API_KEY" \  -H "Content-Type: application/json" \  -d '{"query": "How do vector databases index embeddings?", "depth": "deep", "citationFormat": "apa"}' # Long ones as a job: add "async": true, then GET /v1/jobs/{id}

from tavily import TavilyClient client = TavilyClient(api_key="cf_live_...", api_base_url="https://api.contextfetch.dev/tavily")task = client.research(input="How do vector databases index embeddings?", model="pro")# then GET /tavily/research/{request_id}: content (the report) and sources

Priced like a search, not like a consultant

List price per 1,000 requests, as of 2026-10-17.

List price in US dollars per 1,000 deep research and specialised requests, as of 2026-10-17
Group	Exa Agent (high)	You.com Research (exhaustive)	Parallel Task (ultra)	Linkup Research (S)	Tavily Research (pro)	Valyu DeepResearch (fast)	Google Search grounding	Exa Deep search	ContextFetch deep research (MCP)	Perplexity Sonar Pro	Exa Answer
Per 1,000 requests	$500	$450	$300	$250	$120 (from)	$100	$14	$12 (to $15)	$10 (sub-topics, hub pages, cited report)	$6 (+ tokens)	$5

Pricing

Deep research for your agents

$1.00/ 1k searches

Deep research is 10 credits a report; searches $1.00 per 1,000. Credits never expire. One-time payment. Credits never expire. Full Markdown and pruning included.

Buy 15,000 for $15 All pricing

Questions

What does a deep research call do?

It runs a deep search of the question, searches the topic's sub-topics (the headings of its Wikipedia article, else overview, history, how it works, criticism and latest news), follows the best links of list and index pages among the results, ranks everything with a cross-encoder and builds a report: a summary of the best passages, key facts with dates and numbers, a section per sub-topic and the sources.

Is anything generated?

No. Every sentence in the report is quoted from a source page and cited as [n]. There is no LLM in the pipeline, so there is nothing to hallucinate; your agent writes the prose if it wants prose.

How good is it?

On 40 questions from Google's FRAMES multi-hop benchmark, deep research returned 39.2% of the Wikipedia articles each question needs (standard research 27.6%), and the answer appeared in what it returned for 43.6% (30%). Retrieval only: no LLM answered or graded anything.

What does it cost?

10 credits ($0.01) a report, flat, on the base pack. Standard research is 1 credit per sub-query. Add outputSchema for JSON fields at 1 credit more.