Skip to content

New Official SDKs for TypeScript, Python and Go

SerpKite
Get API key

About

Google results for the software that reads them

SerpKite is a small, remote team building a Google SERP API for AI agents and the developers who build them. We think search data should be cheap, correct and honestly billed.

Why we exist

What we saw

Agents and apps need fresh web results, and Google is where people look. But the options were split: cheap APIs with thin features and credits that expire, or rich APIs at 10–25× the price. Meanwhile the official routes closed: Bing's Search API was retired in August 2025, and Google's Custom Search JSON API shuts down on January 1, 2027.

Google itself kept changing the page: JavaScript became required, num=100 disappeared, and result links became encoded redirects. Each change broke naive scrapers without an error message. We built SerpKite to absorb that churn so you don't have to, and to price it fairly: per page, no expiry, nothing billed for failures.

Principles

How we work

01

Pay for what you use. Ever.

Credits never expire. Failed, empty and blocked requests are free. Every response tells you what it cost.

02

Correct before clever

Resolved URLs, empty-result detection and a parse-quality flag on every response. Silent failures are bugs.

03

Built for the reader that's a model

Markdown, compact JSON and field projection, because agents pay for every token they read.

04

No lock-in

Published OpenAPI spec, MIT-licensed SDKs, no subscription and credits that never expire. Apps leaving Google CSE keep their parsers via /customsearch/v1.

05

Private by default

We never log your query text. Results are cached for a few hours under a hashed key; usage logs keep metadata only, for 31 days.

06

Say what we do

We publish how we collect data, what we don't do yet, and where competitors are the better choice.

Where we are

Honest status

  • Google web search and 12 verticals are live. Bing, DuckDuckGo, Brave and Yandex are planned for Q1 2027.
  • We publish live latency and uptime on status and will publish a reproducible benchmark monthly on /benchmarks.
  • SOC 2 is planned; today we never log query text, and we offer a DPA.
  • We collect logged-out public results only, through proxies. Details in how we collect data.

Product news is on the blog and the changelog.

Start building

Get your API key in 30 seconds

2,500 free credits, then 1,000 every month. No credit card.