Tavily Extract compatible

The call, before and after

# Before
from tavily import TavilyClient
client = TavilyClient(api_key="tvly-...")

# After
client = TavilyClient(api_key=SPECSTREAM_KEY, api_base_url="https://api.specstream.ai/tavily")

reply = client.extract(urls=["https://en.wikipedia.org/wiki/Web_scraping"])
for page in reply["results"]:
    print(page["url"], page["raw_content"][:300])

tavily-python 0.8.5 joins api_base_url to /extract, so extract and search both reach SpecStream. A call with an empty balance answers 432, which the client raises as ForbiddenError, as it does on Tavily.

Apps and SDKs that take the address as a setting

Each line was run on production on November 10, 2026 through that client's own request and reply code. Put your SpecStream key wherever the client takes its key.

LangChain, langchain-tavily 0.2.18: pass api_base_url="https://api.specstream.ai/tavily" to TavilyExtract.

Open WebUI: set WEB_LOADER_ENGINE=tavily and TAVILY_API_BASE_URL=https://api.specstream.ai/tavily in the environment. This one has no field in the admin screen.

LibreChat: set TAVILY_EXTRACT_URL (config key tavilyExtractUrl) to the whole address, https://api.specstream.ai/tavily/extract.

OpenClaw: set plugins.entries.tavily.config.webSearch.baseUrl, or TAVILY_BASE_URL, to https://api.specstream.ai/tavily.

What the route reads, and what it refuses

Read and applied: urls as one string or a list of up to 20, with a host that does not resolve listed in failed_results, and format as markdown or text. The key is read from the Bearer header or from api_key in the body.

Refused with 400: 21 URLs or more, and an empty list. A wrong key gets 401. Every parameter was measured on production on Nov 10, 2026 with tavily-python 0.8.5, and the full pass table lists each one and its result.

Every field that comes back

The reply has results[], each with url and raw_content, failed_results[], each with url and error, response_time in seconds and request_id. The route does not return usage, or a result's images and favicon.

One field is added: specstream, with the call's price_usd. A parser that reads Tavily's fields never sees it.

$0.80 per 1,000 pages, with no credits

Tavily counts 1 credit for every 5 successful URLs on basic and 2 on advanced, at $0.008 a credit on pay as you go. On SpecStream every page that gives text costs $0.0008, and a page with no text is not charged. A busy call gets 429 with Retry-After and charges nothing.

0.908 F1 on WCXB, and the search scores on the same key

The Extract API that answers this route scores 0.908 F1 on WCXB. The Search API on the same key scores 97.9 F-score on all 1,000 official SimpleQA Verified questions, 978 of them correct, and 64.0 item F1 and 43.4 row F1 on all 200 official WideSearch tasks. Every score and its method is on the benchmarks page.

You decide what is kept, on each key

Create the key with zero_retention and a page read made with it keeps no request text and no answer after the reply. spend_cap_cents is the most the key can spend.

curl -X POST https://api.specstream.ai/v1/account/keys \
  -H "Authorization: Bearer $SPECSTREAM_API_KEY" -H "Content-Type: application/json" \
  -d '{"name": "production", "scopes": ["account:read", "finds:spend"], "spend_cap_cents": 500000, "zero_retention": true}'

Search, pages and rows on one key

The same key searches the live web with the Search API at $5 per 1,000 searches of up to 10 results. Deep Search returns every match for a request in plain words as structured rows, at 5¢, 15¢ or $1 a row, with a source URL on every value. The most a run can cost is approved before it starts. The full comparison is on the Tavily alternative page.

Exa Contents

Page text from Exa at $1 per 1,000 pages for each content type you ask for, in Exa's own request and reply. SpecStream reads 1,000 pages for $0.80 and answers the Tavily call your code already sends.

Firecrawl

A scrape API at 1 credit per page, with the Hobby plan at $19 a month and 1,000 free credits a month. SpecStream needs no plan and keeps Tavily's reply shape.

Jina Reader

Page reads billed in tokens by content length, with 10M free tokens on a new key and no dollar rate on its Reader page. SpecStream charges $0.0008 a page, whatever its length.