SERPdive TypeScript SDK

The official TypeScript/JavaScript client for SERPdive, the AI Search API: ask a question, get answer-ready web content that is extracted, cleaned, and sized for an LLM. On a public, replayable 1,000-question benchmark, SERPdive runs at the same speed as Tavily, feeds your LLM 20.2% fewer tokens, and wins 60.7% of decided quality duels. If you are evaluating Tavily alternatives, that benchmark is public and replayable end to end: same questions, same judge, your machine.

There is a free tier, and it has no ceiling. The krill model is free and unlimited under fair use — no card, no credits, nothing to decrement. It returns the shortest set of sentences that still answers (about 700 tokens a search, roughly half what the usual alternatives send), one request at a time, at low priority. Use it to build; switch one word to mako when you need depth and steady latency.

Zero dependencies. Works in Node 18+, Bun, Deno, and edge runtimes (anywhere fetch exists). Full TypeScript types included.

Install

npm install serpdive

Quickstart

import { SerpDive } from "serpdive";

const client = new SerpDive({ apiKey: "sd_live_..." }); // or set SERPDIVE_API_KEY
const response = await client.search("who won the 2026 champions league final", {
  answer: true,
});

console.log(response.answer);
for (const result of response.results) {
  console.log(result.url, result.content.slice(0, 100));
}

Get your API key at serpdive.com/dashboard/keys.

Usage

Choosing a model

// mako (default): answers in a few seconds
await client.search("best rust web frameworks");

// krill: free and unlimited (fair use) — the smallest useful payload
await client.search("who is the ceo of vercel", { model: "krill" });

// moby: reads whole pages, for deep research
await client.search("timeline of the OpenAI board dispute", { model: "moby", answer: true });

model	cost	payload	notes
krill	free, unlimited (fair use)	~700 tokens	one request at a time, low priority, no answer
mako	1 credit	~1k tokens	the default; best for agents and RAG
moby	1.5 credits	up to ~15k tokens	whole page content, cited answer

Options

await client.search("your question", {
  model: "moby",     // "mako" (default), "krill" (free, unlimited), "moby" (whole pages)
  answer: true,      // also return a written answer built from the sources (not on krill)
  maxResults: 5,     // hard cap on delivered results, 1 to 10
});

Localization is automatic: the language of the query picks where we search. There is no country parameter to configure.

Errors

Every API error is a typed exception with a stable code, the human message, and the HTTP statusCode:

import { SerpDive, RateLimitError, QuotaExceededError, SerpDiveError } from "serpdive";

try {
  const response = await client.search("your question");
} catch (err) {
  if (err instanceof RateLimitError) {
    // slow down, then retry
  } else if (err instanceof QuotaExceededError) {
    // monthly credits exhausted
  } else if (err instanceof SerpDiveError) {
    console.error(err.code, err.message);
  }
}

Transient failures (HTTP 502/503) are retried automatically; failed searches are never billed. Tune with new SerpDive({ maxRetries, timeoutMs }).

Response shape

search() resolves to a SearchResponse:

Field	Type	Notes
query	string	your query, echoed
results	SearchResult[]	each has url, content, title, optional ISO date
answer	string | null	only when answer: true was requested
extra_info	ExtraInfo	direct-answer block (weather, rates, scores...) when the query has one
model	string	which model answered
response_time_ms	number	end-to-end latency

Docs

Full documentation: serpdive.com/docs. The API is also self-describing for agents: llms.txt, openapi.json.

License

MIT