browserfoundry.mdtext/markdown · 134 lines · 5,662 bytes · 0 scripts needed/llms.txt
1# browserfoundry
2
3> Search, browser and data APIs for AI agents, behind one key.
4> A Keirolabs product.
5
6- base_url: https://api.keirolabs.cloud
7- auth: `Authorization: Bearer keiro_...`
8- mcp_server: https://keirolabs.space/mcp
9- get_a_key: https://platform.keirolabs.cloud/login
10- free_tier: 750 credits / month, no credit card
11- docs: https://platform.keirolabs.cloud/docs
12
13## Search APIs
14
15Live results, cited answers and deep research from one index of the open web.
16
17| endpoint | does | credits | latency |
18|-----------------------------|-------------------------------------------|---------|-----------|
19| POST /api/v2/search/fast | ranked results | 1 | ~1s |
20| POST /api/v2/search/content | results + full markdown + embeddings | 3 | ~3s |
21| POST /api/v2/answer | a cited answer in one call | 5 | ~2-5s |
22| POST /api/v2/agentic | multi-step research on hard questions | 20 | 30s-3min |
23
24## Browser APIs
25
26Real Chrome sessions your agent can drive when a search result isn't enough.
27
28- POST /v2/browsers/lease: a live session in under a minute
29- stealth by default: a real Chrome fingerprint
30- residential exits: a fresh IP for every session
31- billing: per minute, only while the session is alive
32
33## Data APIs
34
35Clean, structured content from any page, one URL or ten thousand at a time.
36
37| endpoint | does | credits |
38|-----------------------------|-------------------------------------------|----------------|
39| POST /api/v2/extract | any URL to clean markdown | 3 |
40| POST /api/v2/search/batch | up to 10,000 queries per job | 1 per query |
41| Google SERP | full results pages as JSON | see docs |
42
43## Quickstart
44
45```bash
46curl -X POST https://api.keirolabs.cloud/api/v2/search/fast \
47 -H "Authorization: Bearer $KEIRO_API_KEY" \
48 -H "Content-Type: application/json" \
49 -d '{"query":"latest open source LLMs","maxResults":5}'
50```
51
52## Response shape: /search/fast
53
54```json
55{
56 "results": [
57 { "title": "…", "url": "https://…", "snippet": "…", "score": 0.97, "published_date": "…" }
58 ],
59 "total_results": 5,
60 "latency_ms": 942,
61 "credits_used": 1
62}
63```
64
65## Benchmarks
66
67| benchmark | keiro | next best |
68|-----------------------------------|--------|------------------|
69| SimpleQA (deep search) | 95.3% | Firecrawl 94.7% |
70| FinanceBench (accuracy) | 78% | Valyu 73% |
71| AIMultiple Agentic '26 (/20) | 15.20 | Brave 14.89 |
72
73### Deep search on SimpleQA
74
75953 of 1,000 seeded questions (seed 42) had the gold answer in the returned data.
76Retrieval-only: no reader model, no LLM judge. $4.44 per 1,000 requests.
77Snippets alone reach 72.4%; reading pages in full adds 15.3 points (7.9 pages per question).
78
79| provider | SimpleQA |
80|-------------------|----------|
81| Keiro deep search | 95.3% |
82| Firecrawl | 94.7% |
83| Tavily | 93.3% |
84| you.com | 92.09% |
85| Exa | 91.9% |
86| Parallel | 91.0% |
87| Claude Search | 90.5% |
88| Perplexity | 85.92% |
89| Google | 82.15% |
90| Brave Search | 76.05% |
91
92Best published run per provider, Sept 2026. The others use a reader model and a
93judge, so the gaps are directional. Report: https://keirolabs.cloud/Deep-search
94
95## Works with
96
97Claude Code, Cursor, Codex, Windsurf, OpenCode, Ollama, LangChain, LangGraph,
98Vercel AI SDK, CrewAI, n8n, Zapier. One API key and one MCP server.
99
100## Pricing
101
102| plan | monthly | yearly | credits / month | rate limit |
103|---|---|---|---|---|
104| Free | $0 | $0 | 750 | 15 req/min on lite search |
105| Essential | $29.99 | $300/yr, save $59.88 | 11,250 | 60 req/min on lite search |
106| Pro | $49.99 | $504/yr, save $95.88 | 33,750 | 300 req/min on lite search |
107| Startup | $99.99 | $996/yr, save $203.88 | 112,500 | 600 req/min on lite search |
108
109/search/lite costs 0.0375 credit per request on paid plans ($0.10 per 1,000 on Essential), 1 on Free.
110Paid plans pay half on fast, content, extract and answer. Batch is free on Startup.
111One-time pack credits expire 6 months after purchase.
112One-time credit packs from $10. All plans: https://keirolabs.cloud/pricing
113
114## Data policy
115
116- never sold: queries and customer data are not sold, shared, or brokered
117- never leaves the platform: processed in-region on our own infrastructure
118- never trained on: your queries don't train models, ours or anyone's
119- pinned to your region: one X-Region header keeps data in EU, US, or APAC
120
121## Blog (each post also as markdown at /blog/<slug>.md)
122
123- [Google Custom Search API is shutting down: what to use instead](https://www.browserfoundry.cloud/blog/google-custom-search-api-alternative.md)
124- [DuckDuckGo search API: free, limits, and what to use instead](https://www.browserfoundry.cloud/blog/duckduckgo-search-api.md)
125- [URL to markdown for LLMs: 7 ways compared](https://www.browserfoundry.cloud/blog/url-to-markdown-for-llm.md)
126- [What web search costs in the OpenAI, Claude, Gemini and Perplexity APIs](https://www.browserfoundry.cloud/blog/llm-web-search-api-pricing.md)
127- [Browserbase pricing explained, and cheaper cloud browsers](https://www.browserfoundry.cloud/blog/browserbase-pricing-alternatives.md)
128
129## Links
130
131- docs: https://platform.keirolabs.cloud/docs
132- status: https://platform.keirolabs.cloud/status
133- blog: https://keirolabs.cloud/blog
134- contact: team@keirolabs.cloud
G24.5on G273NPS
browserfoundry · by Keirolabs

Every Web API
Your Agent Needs.

Search, browsers, extraction and research behind one key. Forged by Keirolabs for agents that work on the live web.

POST /api/v2/search/fast
Try
Start FreeRead the docs750 free credits every month · no credit card

Trusted by engineers from

  • RBS International
  • Stack Overflow
  • Supabase
  • Qdrant
  • Replicate
  • Modal
  • Resend
  • Turso
The call

What You Send.
What You Get Back.

Pick an API, pick a language. One key and one shape of request across all of them.

200 OK · 942 ms · 1 credit
Request
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 open source LLMs","maxResults":5}'
Responsesample values, real shape
{
  "query": "Latest open source LLMs",
  "results": [
    {
      "title": "…",
      "url": "https://…",
      "snippet": "…",
      "score": 0.97,
      "published_date": "2026-09-28T09:12:00Z"
    },
    { "title": "…", "url": "https://…", "score": 0.94 },
    { "title": "…", "url": "https://…", "score": 0.91 }
  ],
  "total_results": 5,
  "latency_ms": 942,
  "credits_used": 1
}
Latency
~1s
Cost
1 credit
Plans
Free and up
Returns
ranked sources + scores
Deep search

The Most Accurate
Web Search API.

Gaugamela, 331 BC. Alexander didn't fight the whole line.
He drove one wedge straight through its centre.

The web is mostly noise. Deep search doesn't hand you ten links into it. It goes straight for the fact, reads every page to the end, and comes back with the proof.

Code & every result on GitHubRead the report →

SimpleQA accuracybest published run per provider · %
  1. Keiro deep search95.3
  2. Firecrawl94.7
  3. Tavily93.3
  4. you.com92.09
  5. Exa91.9
  6. Parallel91.0
  7. Claude Search90.5
  8. Perplexity85.92
  9. Google82.15
  10. Brave Search76.05

Sept 2026. Ours is retrieval-only: the answer has to be in what we return, with no model to rescue it. The others use a reader model and a judge, so read the gaps as directional.

95.3%
of SimpleQA answered953 of 1,000 questions
+15.3
points from reading pages in fullsnippets alone: 72.4%
7.9
pages read per questionend to end, every one
$4.44
per 1,000 requestsone meter for all of it
What we believe

The web is the largest thing humanity has ever written. Until now, only humans could read it.

That is ending. Soon most of its readers will be machines, acting for people, at a speed no person could.

They will need all of it: live, clean, and true enough to act on. We are building the way in.

The foundry

One Key.
Every Way onto the Web.

01 / 03 — Point at anything.

Search APIs

Live results, cited answers and deep research from one index of the open web.

  • POST /api/v2/search/fastRanked results in about a second

  • POST /api/v2/search/contentResults plus full markdown and embeddings

  • POST /api/v2/answerA cited answer in one call

  • POST /api/v2/agenticMulti-step research on hard questions

02 / 03 — Reach in and click.

Browser APIs

Real Chrome sessions your agent can drive when a search result isn't enough.

  • POST /v2/browsers/leaseA live session in under a minute

  • on every sessionA real Chrome fingerprint

  • on every sessionA fresh IP for every session

  • meteredPay only while the session is alive

03 / 03 — Bring it all back.

Data APIs

Clean, structured content from any page, one URL or ten thousand at a time.

  • POST /api/v2/extractAny URL to clean markdown

  • POST /api/v2/search/batchUp to 75 queries per job

  • full results pagesGoogle's results, structured as JSON

  • keirolabs.space/mcpEvery endpoint as an agent tool

Works with your stack

Plugs into
Every Agent You Run.

One API key and one MCP server. The same spark reaches every tool on the ring.

one key · one MCP server

Benchmarks

Don't Take Our Word.
Take the Numbers.

Public benchmarks, on the kinds of requests agents actually make.

FinanceBench · Patronus 150-question split · accuracy
Financial document QA: FinanceBench · Patronus 150-question split · accuracy
Keiro78%
Valyu73%
Parallel67%
Exa63%
Google55%

FinanceBench methodology →Keirolabs-API / Benchmarks

The live web,
Finally Working for You.

Stop babysitting proxies and headless fleets. browserfoundry fetches, renders and cleans every page on our side.

Endpoints · credits per request
  • Fast search/search/fast0.5
  • Search + content/search/content1.5
  • Answer/answer1
  • Research/agentic20
Your data

Your Data Never
Becomes the Product.

What you search is yours. The whole policy, in four lines.

  • Never soldWe don't sell, share, or broker your queries or customer data. Ever.
  • Never leaves the platformRequests are processed in-region on our own infrastructure. No third-party data pipelines.
  • Never trained onYour queries don't train models. Not ours, not anyone's.
  • Pinned to your regionOne X-Region header keeps your data in the EU, the US, or APAC.
Pricing

Simple Pricing,
no Surprises.

Pay in credits for what you actually call. Start free and move up when your agents do.

Free
$0/mofree, forever

Start with the basics. Test every endpoint.

750
credits / month
750
searches on /search/lite
  • 15 req/min on lite search
  • Search, content, extract, answer, deep
  • Community support
  • No credit card required
Start Free
EssentialMost popular
$29.99/mobilled monthly · $25/mo if yearly

The perfect starting point for serious projects.

11,250
credits / month
300,000
searches on /search/lite
  • 60 req/min on lite search
  • Half price on fast, content, extract, answer
  • Batch at 0.5 credit / query
  • Drift + Remem included
  • Email support
Start at $29.99
Pro
$49.99/mobilled monthly · $42/mo if yearly

More power, more speed, more control.

33,750
credits / month
900,000
searches on /search/lite
  • 300 req/min on lite search
  • Priority queue (2x)
  • Webhook integrations
  • Priority support
Go Pro
Startup
$99.99/mobilled monthly · $83/mo if yearly

Built for teams that need scale.

112,500
credits / month
3,000,000
searches on /search/lite
  • 600 req/min on lite search
  • Batch · free and unlimited
  • Priority infrastructure
  • Dedicated account manager
Go Startup

Not sure which one? Run your numbers and compare every plan →

Start with browserfoundry

Ready to Give Your
Agent the Web?

Start free today. 750 credits every month, no credit card, first result in under two minutes.

Start for Free