The web search API for RAG
A model can only answer from what retrieval hands it. SpecStream leads WideSearch (60.9, on the low-cost model; the best other API scores 57.6 even on the larger model) and SimpleQA Verified (99). See the benchmarks. So the pages your model reads are the right ones more often, at $5 per 1,000 searches. One call returns ranked sources with snippets and publish dates, and passages=true adds the text of the top pages. The Extract API reads any page as markdown for $0.80 per 1,000, falls back to the site's feed or an archive copy when the live page is blocked, and charges nothing for a page with no text. SpecStream is the live web data engine for teams and agents.
New accounts get $10 of credit for their first 30 days, no card.
Six web search APIs for RAG, side by side
| SpecStream | Exa | Tavily | Parallel | Perplexity | You.com | |
|---|---|---|---|---|---|---|
| WideSearch, low-cost model | 60.9 | 53 | 47.9 | 45.5 (Fast), 44 (Turbo) | 47 | Not in the published scores |
| SimpleQA Verified, larger model | 99 | 91 | 92 | 97 (Advanced, Basic) | 95 | Not in the published scores |
| Price per 1,000 searches | $5 | $7 (auto, fast), $4 (lite, instant) | $8 (basic), $16 (advanced) | $5 (basic, advanced), $1 (turbo, fast) | $5, $1 (Fast Search) | $5 |
| Page text in the search reply | passages=true, about 1 s more | Not stated | Not stated | Not stated | Not stated | Not stated |
| Price per 1,000 pages read | $0.80 | $1 (Contents) | $1.60 (1 credit per 5 URLs) | $1 (Extract) | $0.50 (fetch_url in the Agent API) | $1 (Contents) |
| A page with no text | Not charged | Not stated | Not stated | Not stated | Not stated | Not stated |
| When the live page is blocked | Browser-like retry, a real browser, the site's feed, then an archive copy, labeled in read_from | Not stated | Not stated | Not stated | Not stated | Not stated |
| Read time on each value | seen_at on every value, published on each source | Not stated | Not stated | Not stated | Not stated | Not stated |
| Work beyond one search | Deep Search: structured rows at 5¢, 15¢ or $1 a delivered row | Agent, $0.012 to $1.00 a request | Research, 4 to 250 credits a request | Task API, $5 to $300 per 1,000 runs | Agent API, no price per effort stated | Research, $12 to $1,200 per 1,000 calls |
Why RAG pipelines retrieve with SpecStream
The right passages, more often
When retrieval misses, the model writes a fluent answer from the wrong page. On SimpleQA Verified, SpecStream scored 99 on the larger model and 96 on the low-cost one. On WideSearch, which asks for many facts spread across many pages, it scored 60.9 on the low-cost model, above every other API in the published scores on either model.
Context that carries its source
Each item in sources has its url, title, site, snippet, and published date. The answer and each value in rows carry the page they were read on and seen_at, the time SpecStream read it. Your prompt can tell the model where a fact came from and how old it is, and your app can show the reader the same.
Source URL and read time on every value.
Whole pages, and no charge for empty ones
POST /v1/extract reads 1 to 20 URLs per call as markdown or text. When a site blocks the first read, SpecStream retries like a browser, then with a real browser, then from the site's feed, then from an archive copy, and read_from says which one you got. When the job grows into a dataset, Deep Search builds structured rows at 5¢, 15¢ or $1 a row, plus 5¢ for an email and 15¢ for a phone, and keeps them up to date on a schedule.
How SpecStream scores
The same numbers as the benchmarks page, from the official tasks and scorers.
| SpecStream |  Parallel |  Exa |  Tavily |  Perplexity | |
|---|---|---|---|---|---|
| Web researchWideSearch, GPT-5.6 Luna | 60.9 | 45.5 | 53.0 | 47.9 | 47.0 |
| Factual accuracySimpleQA Verified, GPT-5.6 Luna | 96 | 94 | 91 | 94 | 94 |
| Factual accuracySimpleQA Verified, GPT-5.6 Sol | 99 | 97 | 91 | 92 | 95 |
How to start
New accounts get $10 of credit for their first 30 days, no card. Create an API key and point your retriever at GET /v1/web-search, with passages=true when you want page text in the same reply.
To read pages you already know, send them to the Extract API. Guides for LangChain, LlamaIndex, CrewAI, the OpenAI API, and the Vercel AI SDK are in the framework docs.
curl https://api.specstream.ai/v1/extract \
-H "Authorization: Bearer $SPECSTREAM_API_KEY" \
-H "Content-Type: application/json" \
-d '{"urls": ["https://en.wikipedia.org/wiki/Narbonne"], "format": "markdown"}'Frequently asked questions
Something we didn't answer? Talk to a human
What is the best web search API for RAG?
passages=true adds the text of the top pages.What does SpecStream cost in a RAG pipeline?
What format do pages come back in?
read_from says whether the text came from the live page, the site, or an archive copy, with captured_at for an archive read.Is there a lower price per page read?
fetch_url at $0.50 per 1,000 in its Agent API. SpecStream reads pages at $0.80 per 1,000, charges nothing for a page with no text, and falls back to the site's feed or an archive copy when the live page is blocked, so a blocked page can still come back with text.Does SpecStream work with LangChain and LlamaIndex?
https://api.specstream.ai/mcp.Sources
SpecStream wrote this page. Each rival figure comes from that company's own page.
- SpecStream pricingspecstream.ai/pricing, read Nov 8, 2026
- SpecStream for agentsspecstream.ai/agents.md, read Nov 8, 2026
- Other providers' benchmark scores, Sep 9, 2026parallel.ai/benchmarks, read Nov 8, 2026
- Exa pricingexa.ai/pricing, read Nov 8, 2026
- Tavily pricingtavily.com/pricing, read Nov 8, 2026
- Tavily API creditsdocs.tavily.com/documentation/api-credits, read Nov 8, 2026
- Parallel pricingdocs.parallel.ai/getting-started/pricing, read Nov 8, 2026
- Perplexity pricingdocs.perplexity.ai/getting-started/pricing, read Nov 8, 2026
- You.com pricingyou.com/pricing, read Nov 8, 2026