Real browsers · Real iOS · Real Android · p50 < 3s cold start

Cloud emulators,built for agents.

Fast, reliable browser, iOS, and Android emulators you can see, control, and debug remotely — driven by your Playwright suite, your CI, or your AI agent. Half the price of Sauce Labs and BrowserStack, an order of magnitude faster to start.

Start free

3,500+ device configurations•Live remote control•p50 < 3s cold start

BrowseriOSAndroidCanvasCodeSplit

● Compiled · Playwright

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

// tests/gmail-login.spec.ts
import { test, expect } from '@kifas/playwright';

test('gmail login with TOTP', async ({ kifas, page }) => {
  // Spawn a warm Chromium on Kifas · p50 2.9s
  const browser = await kifas.browser.spawn('chromium-131');

  // One-line authenticated session — Auth Catalog
  await kifas.auth.loginAs('gmail', {
    user: 'qa@acme.dev',
    totp: process.env.KIFAS_TOTP_SECRET,
  });

  // Save state for reuse across 50 tests
  await kifas.state.save('admin_logged_in');

  await page.goto('https://gmail.com/inbox');

  // AI-explored step — first run, deterministic after
  await kifas.ai.step('Compose new email to raj@acme.dev');

  await expect(page.getByText('Welcome, Maya')).toBeVisible();
});

Maya needs

Gmail login with 2FA in under 15 lines

Raj needs

Enterprise checkout regression at 120 parallels

Jonah needs

Agent eval with persistent multi-turn state

You can

Click a node to inspect it live

⬢

TOTP provider

⬢

claude-sonnet

Load SDK

@kifas/sdk · v1.4.2

🔒 locked 234ms

Spawn browser

Chromium 131 · Linux

🔒 locked 207ms

user: qa@acme.dev + TOTP

🔒 locked 180ms

Save state

handle: admin_logged_in

🔒 locked 286ms

Navigate

gmail.com/inbox

🔒 locked 449ms

Vibe-step

"Compose new email"

◉ AI 160ms

Text contains

"Welcome, Maya"

🔒 locked 302ms

Slack notify

#qa-fails on fail

manual 151ms

60%

Integrates withPlaywrightSeleniumCypressClaude CodeCursorWindsurfGitHub ActionsJira

Seven differentiators

Every feature, callable by agents.

Kifas was built MCP-first — every product surface is a first-class tool. The dashboard is a UI on top of the same API your agents use.

01 / MCP-NATIVE

One canonical endpoint, 40+ tools

Streamable HTTP, OAuth 2.1, Client ID Metadata Documents, async Tasks. ~200–400 token snapshots. Published at .well-known/mcp-server-card.

STDIO · HTTP · OAuth 2.1

02 / SPEED

p50 < 3s cold start, published weekly

Warm pools on Graviton spot. Open-source benchmark harness against Sauce, BrowserStack, Browserbase, Device Farm. No queue, no tail.

3.1s p50 · 7.4s p95

03 / AUTH CATALOG

Gmail, TOTP, magic links — one line

kifas.auth.login_as("google_workspace") — SSO, 2FA, email OTP, SMS, CAPTCHA bypass. No other testing vendor ships this.

Google · Okta · Microsoft · TOTP

04 / VIBE TESTING

Describe it. Kifas builds the graph.

Server-side agent drives a cloud browser with the full Kifas palette, emits a declarative test that is simultaneously NL, a workflow, and Playwright code.

05 / HYBRID ENGINE

AI explores. Playwright executes.

First run explores. Every re-run hits compiled Playwright — seconds, not minutes. On break, auto-reheals with a diff you approve in CI.

🔒 deterministic by default

06 / MANAGED STATE

Named, test-shaped state handles

Cookies, localStorage, IndexedDB, service workers — saved under a name, reused across sessions, auto-refreshed on expiry. Collapses suite time 30–50%.

07 / VISUAL WORKFLOW BUILDER

Tests that a QA manager can read. Sub-workflows that 50 tests can share.

A node-and-edge canvas where every test lives as three synced representations — natural language, a visual graph, and deterministic code. Agents edit the graph through MCP. Non-engineers finally audit coverage. No testing vendor ships this today. It's our biggest wedge.

vs. the Selenium-era incumbents

Half the price. An order of magnitude more usable.

Sauce Labs

BrowserStack

Kifas

5 parallels, monthly list

$745

$645

$299

Session cold start (p50)

30–60s

~15s

< 3s

MCP server spec coverage

basic

subset

✓ full, +Tasks, +Elicitation

One-line SSO / 2FA login

bring your own

✓ Auth Catalog

Named, persistent state handles

10s cacheId only

—

✓ per branch, per env

Visual workflow builder

—

✓ n8n-style, agent-editable

Hybrid AI-explore → deterministic compile

—

✓ auto-reheal + diff

Unlimited seats on every paid tier

per-seat

✓

MCP-native, from day one

Your Claude Code, Cursor, or Windsurf agent just gets it.

Install the server once. Your agent can now author workflows, log into Gmail, save state, run a suite against 120 browsers, and read the results — with one consistent tool namespace.

Read the MCP card

~/acme-web · claude-code

// Install Kifas’s MCP server
$ npx @kifas/mcp@latest --install cursor
→ Installed 42 tools. Authenticated as maya@acme.dev

claude > "Test our checkout with a new Gmail user,
apply promo GUMBO20, verify the total."

→ Kifas: workflow.create("checkout-promo")
→ Kifas: workflow.add_node(type: "auth", primitive: "gmail")
→ Kifas: workflow.add_node(type: "action", ai_explore: true)
→ Kifas: workflow.run() · 14 nodes · 3.1s start · passed

// Compiled to /tests/checkout-promo.spec.ts

Pricing

One SKU. No sprawl. Margins that don't lie.

Compute minutes and model access in one bucket. Lite and Pro models bundled as monthly credit on every paid tier. Premium frontier models pass through at 1.4× at-cost — use BYOK on Business+ to skip our markup entirely.

L1 · Lite

Haiku 4.5 · Gemini Flash · gpt-4o-mini

≈ $0.0008 / vibe-step · runs locally on warm pool

L2 · Pro

Sonnet 4.5 · Gemini 2.5 Pro · gpt-5-mini

≈ $0.012 / vibe-step · best for healing & explore

L3 · Premium

Opus 4.5 · gpt-5 · o4-deep-research

≈ $0.18 / vibe-step · gnarly multi-page flows only

BYOK

Your Anthropic / OpenAI / Bedrock key

Zero markup · we pass through at cost

Hobby

Kick the tires. Build one workflow end-to-end.

Compute

Browser300 min

Android emulator—

iOS simulator—

Model access

L1Lite models$10 credit

L2Pro models—

L3Premium models—

·Bring your own key—

1 parallel · 3 state handles · 3 workflows
MCP server · Chromium / Firefox / WebKit

Start free

Starter

$99 /mo

Side projects and small suites. ~23% gross margin at 85% utilization.

Compute

Browser3,500 min

Android emulator600 min

iOS simulator150 min

Model access

L1Lite models$25 credit

L2Pro models$15 credit

L3Premium modelspay-as-you-go

·Bring your own key—

2 parallels · 25 states · 25 workflows
Visual regression · Analytics · Unlimited seats

Choose Starter

Team

$499 /mo

The scale-up default. ~30–40% cheaper than BrowserStack at the same volume.

Compute

Browser25,000 min

Android emulator5,000 min

iOS simulator400 min

Model access

L1Lite models$100 credit

L2Pro models$75 credit

L3Premium modelspay-as-you-go

·Bring your own key—

5 parallels · 100 states · 200 workflows
Auth Catalog · Hybrid engine · Kifas Tunnel

Choose Team

Business

$1,499 /mo

Org-wide test estates. Brings BYOK so your AI bill stays predictable.

Compute

Browser90,000 min

Android emulator18,000 min

iOS simulator2,000 min

Model access

L1Lite models$200 credit

L2Pro models$250 credit

L3Premium models$50 credit

✓Bring your own keyZero markup

15 parallels · Unlimited workflows
iOS Simulators · SSO / SAML · Audit logs

Talk to sales

Overages, metered — priced to actually be sustainableHard cap or auto-scale, your choice. Bills never surprise.

Browser-minute

$0.012 /min

~3× our COGS · 65% margin

Android emulator-minute

$0.04 /min

ARM-native on Graviton metal

iOS simulator-minute

$0.08 /min

EC2 Mac M4 · genuinely expensive

Real device-minute

$0.30 /min

racked iPhones & Pixels · finite

L1 Lite tokens

1.4× at-cost

credit included · cheap to begin with

L2 Pro tokens

1.4× at-cost

$1.4 per $1 of provider spend

L3 Premium tokens

1.2× at-cost

thin margin · use BYOK to skip

BYOK (Business+)

$0 markup

your provider invoices you direct

Enterprise from ~$60K/yr · SOC 2 · Private cloud · BYOK + Bedrock / Vertex routing · Dedicated TAM · Real iOS fleet · See all features

Ship tests that your whole team can read.

Free tier, no card, generous MCP quota. The first green run takes under 3 minutes.

Start free Book a demo