Migrate from Exa

The call, before and after

# Before
curl -X POST https://api.exa.ai/search \
  -H "x-api-key: $EXA_API_KEY" -H "Content-Type: application/json" \
  -d '{"query": "open source vector databases", "numResults": 10}'

# After
curl -X POST https://api.specstream.ai/exa/search \
  -H "x-api-key: $SPECSTREAM_API_KEY" -H "Content-Type: application/json" \
  -d '{"query": "open source vector databases", "numResults": 10}'

The body, the header name and the parser stay. The base address gains /exa, and the key changes. Code that tracks spend with costDollars.total keeps working, because the field holds the price of the call.

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-exa 1.1.0: set exa_base_url="https://api.specstream.ai/exa" on ExaSearchRetriever.

Every parameter, and what the route does with it

Read and used in the search: query, numResults (1 to 100), includeDomains and excludeDomains, includeText and excludeText (searched as exact phrases), and startPublishedDate and endPublishedDate.

Read to choose the search: a type that starts with deep, contents.text and a numResults above 10 run the default search; every other call runs the fast search. The price is the same on both.

Read to shape the reply: contents.text (true, or an object with maxCharacters) and contents.highlights. Both also work at the top of the body, as older clients send them. Accepted and not used: category, userLocation, contents.summary, livecrawl and the rest, so a client that sends them is never refused.

Every field that comes back

The reply has requestId, results, searchTime in milliseconds and costDollars with total. Each result has id, which is its URL, url, title, publishedDate and author. publishedDate is null when the page has no date, and author is always null.

With contents.text, each result carries text: the text of its page closest to your query, up to about 3,500 characters, cut to maxCharacters when you send it. With contents.highlights, it carries highlights, a list with the result's snippet. The route does not return score, summary, image or favicon.

One field is added: specstream, with what the Search API returns beyond Exa's fields. A parser that reads Exa's fields never sees it.

$5 per 1,000 searches, and 100 results for $0.059

On Exa 1,000 auto searches of 10 results cost $7, every result above 10 adds $1 per 1,000 results, and page text is $1 per 1,000 pages per content type. On SpecStream a call of up to 10 results costs $0.005 with the page text included, and each started block of 10 more adds $0.006, so 100 results cost $0.059 against $0.097 on auto. A call stopped by a 429 charges nothing.

The most accurate web search API, in about one second

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

An Exa call of up to 10 results without page text or a deep type runs the fast search. Measured on production on November 9, 2026, its median was 0.82 s at 1 search a second and 1.07 s at 20 a second, the 95th percentile stayed between 1.09 and 1.73 s, and it answered 718 of 720. It takes up to 8,800 searches a minute on one account.

You decide what is kept, on each key

Create the key with zero_retention and a search made with it keeps no request text and no answer after the reply: it is not in the history, its answer is not reused for another call, and its log lines carry neither. Every account can set it. 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 reads up to 20 pages or documents a call with the Extract API at $0.80 per 1,000 pages, and a page with no text is not charged. 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 Exa alternative page.

Tavily

A search API billed in credits: $0.008 a credit pay as you go, 1 credit for a basic search and 2 for an advanced one, up to 20 results a call. SpecStream returns up to 100 results in one Exa call, for $0.059.

Brave Search API

A search API on Brave's own index at $5 per 1,000 requests, with $5 in free credits every month and 50 queries per second, in Brave's own reply shape. SpecStream costs the same per call and answers in Exa's shape.

Parallel

A Search API that starts at $1 per 1,000 requests, with 600 Search calls a minute on the Developer plan. Its migration guide maps Exa search types to Parallel's own modes. SpecStream answers the Exa call as it is, and takes up to 8,800 searches a minute on one account.

Serper

Google results as JSON, sold in credit packs from $1.00 per 1,000 for 50,000 credits to $0.30 per 1,000 for 12.5 million, valid for 6 months. A top-up on SpecStream never expires, and the call your code sends to Exa keeps working.