WebSearch
APIs
Submit an API
★ KEIRO20 = 20% off Keirolabs Keirolabs $0.03/1kSerper $0.30/1kZenserp $0.90/1kParallel $1.00/1kPerplexity $1.00/1kSearchApi $1.40/1kSerpApi $1.96/1kExa $4.00/1kBrave $5.00/1kLinkup $5.00/1kYou.com $5.00/1kTavily $5.00/1kOpenAI Web $10/1kClaude Web $10/1k Full price index →

Blog/Deep dive

Keirolabs: pricing, endpoints, benchmarks and how to start

The complete guide to Keirolabs' search, browser and data API. Every endpoint and its credit cost, plans from $29.99, what $0.03 per 1,000 search really means, the public SimpleQA results, and the honest caveats.

TL;DR

  • Keirolabs is Keirolabs' API for AI agents: search, page content, answers, deep search, Google SERP, extract and batch on one key and one credit balance.
  • Lite search is $0.10 per 1,000 on Essential ($29.99), $0.06 on Pro and $0.03 on Startup. That's the lowest list price we track.
  • Deep search found the answer for 96.5% of 200 SimpleQA questions, with code and per-question results in a public repo.
  • 750 free credits, no card. Through Black Friday, KEIRO20 takes 20% off, and teams leaving Exa or Tavily get two months free on a yearly plan.
$0.03Per 1,000 lite searches on Startup
96.5%SimpleQA, deep search, public repo
1 keySearch, browser and data APIs
600/minLite rate limit on Startup

Most teams building agents end up with three or four vendors: a search API, a scraper for page text, a SERP API for Google, and a browser service for the pages that fight back. Keirolabs, from Keirolabs, puts all of that behind one API key and one credit balance, and prices the basic search lower than anything else we track.

This is the full guide: what it does, what every call costs, what the benchmarks say, and where it’s not the right fit.

What Keirolabs is

Keirolabs’ APIs come in three families:

FamilyWhat it does
Search APIsRanked results from Keirolabs’ own index, results with full markdown, cited answers, deep search with ranked evidence
Browser renderingExtract steps up to JavaScript rendering (deep and heavy modes) for pages that block plain requests, so you rarely need a separate headless browser
Data APIsAny URL to clean markdown, batch search, and Google results as structured JSON

There’s a remote MCP server that exposes every endpoint as an agent tool, and it works with Claude Code, Cursor, Codex, LangChain, LangGraph, the Vercel AI SDK and n8n.

Pricing

From browserfoundry.cloud/pricing, November 5, 2026.

Plans

PlanMonthlyYearly (per month)Credits/monthLite searches/monthLite rate limit
Free$0—750 (once)—15/min
Essential$29.99$2511,250≈300,00060/min
Pro$49.99$4233,750≈900,000300/min
Startup$99.99$83112,500≈3,000,000600/min
EnterpriseCustomCustomCustomCustomCustom

Pro adds a 2× priority queue and webhooks; Startup adds free, unlimited batch search and a dedicated account manager.

What each call costs

EndpointCreditsPer 1,000 (Essential → Startup)
Lite search /search/lite0.0375$0.10 → $0.03
Fast search /api/v2/search/fast0.5$1.33 → $0.44
Search + markdown /api/v2/search/content1.5$4.00 → $1.33
Extract a URL /api/v2/extract1$2.67 → $0.89
Cited answer /api/v2/answer1$2.67 → $0.89
Deep search /api/v2/deep2$5.33 → $1.78
Google SERP /api/v2/search/google2$5.33 → $1.78
Batch search0.5 per query (free on Startup)—

Retries, pagination and streaming within one request aren’t charged again.

Credit packs (no subscription)

PackPriceCredits
Starter$101,500
Growth$305,000
Scale$10020,000

Packs expire 12 months after your latest top-up. Subscription credits are used first.

What “$0.03 per 1,000” means in practice: one million lite searches a month uses 37,500 credits, a third of the Startup plan’s 112,500. The same million searches costs $1,000 on Parallel fast, $4,000 on Exa lite, $5,000 on Brave and $5,000–8,000 on Tavily.

Benchmarks

Keirolabs publishes its SimpleQA runs with code and per-question results, which most vendors don’t:

RunScoreQuestionsRepo
Deep search, full response96.5%200deep-search-benchmark
Deep search, ~900-token evidence only94.0%200same
Lite + open crawl recipe (top 8 pages)92.0%500lite-search-benchmark
Lite, titles and snippets only59.4%500same

Read these the right way. Keirolabs scores whether the answer appears in what comes back, with an LLM judge for variant forms. Most other vendors have a model answer the question and grade that. The two are close but not identical. And lite snippets alone hold the answer 59.4% of the time, so for factual questions, pair lite with content or use deep. Every claim, side by side: benchmarks.

Your first request

Sign up at platform.keirolabs.cloud, create a key (it starts with keiro_) and call any endpoint with it as a Bearer token:

curl -X POST https://api.keirolabs.cloud/api/v2/search/fast \
  -H "Authorization: Bearer $KEIRO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "latest rust release"}'

Swap the path for /api/v2/search/content to get markdown, /api/v2/answer for a cited answer, or /api/v2/deep for deep search. Same key, same credits.

Where it fits, and where it doesn’t

Great fit

Agents and RAG at volume

High search counts, tight budgets, and a need for page text and answers on one bill.

Great fit

Replacing a vendor stack

Search API + scraper + SERP API + browser service, consolidated.

Think twice

Pure rank tracking

If all you need is Google’s page at huge volume, Serper’s $0.30–1.00 beats a 5-credit SERP call.

Think twice

Semantic “find similar” search

Exa’s neural index is built for that specific job.

The offers right now

  • 750 free credits at signup, no card.
  • Black Friday: code KEIRO20 for 20% off a paid plan.
  • Switching from Exa or Tavily: send your invoice to team@keirolabs.cloud for two months free on any yearly plan.
  • Yearly billing saves up to $204 a year on Startup.

My honest read: for most agent and RAG workloads, Keirolabs is the first API I’d price against, because it covers the most jobs at the lowest per-call cost, and its accuracy claims come with code you can run. Replay a day of your real queries on the free credits before you commit. Start here.

FAQ

Is BrowserFoundry the same as Keirolabs?

BrowserFoundry is the product; Keirolabs is the company. The API lives at api.keirolabs.cloud and keys start with keiro_.

Does Keirolabs have its own index?

Yes. Lite, fast, content and deep search run on Keirolabs’ own index. The Google SERP endpoint returns Google’s results separately.

Is there an MCP server?

Yes, a remote MCP server that exposes every endpoint, for Claude Code, Cursor, Codex and other MCP clients.

What does a Google SERP call cost?

2 credits: $5.33 per 1,000 on Essential, $2.96 on Pro and $1.78 on Startup.

Topic hub

More on search apis for ai and agents

  1. Best web search APIs in 2026, ranked on two benchmarksStart here
  2. Web search in LM Studio and Open WebUI: free and paid setups, step by step
  3. How to search the web from Python: ddgs, googlesearch and 3 APIs (with code)
  4. Tavily MCP, Exa MCP and Keirolabs MCP: setup for Claude Code, Cursor and LM Studio
  5. Tavily vs Perplexity vs Exa: search API pricing, accuracy and which to pick
  6. OpenAI web search API: pricing, limits and when a search API is cheaper
  7. Exa vs Tavily: pricing, index, accuracy and which to pick
  8. Exa API pricing and API key: every endpoint, explained
  9. Tavily API key: get one free in 2 minutes, plus the real cost per search
  10. Web search APIs for AI agents: MCP servers, token costs and what actually matters
  11. Free web search APIs in 2026: every free tier, and how far it really goes
  12. What is a web search API? Types, prices and how to pick one (2026)
  13. Perplexity API pricing (Nov 2026): $5 per 1,000 searches, or $1 on Fast
  14. Tavily alternatives: 8 search APIs for AI agents, compared

Keep reading