Home/Features
Platform capabilities

Everything you need to
automate at scale

Browser-equipped sandboxes for web automation and secure execution, unified under one SDK. Build AI agents, internal ops workflows, and document-heavy automations.

Paired browser

A live browser inside every workflow

Launch a browser-equipped sandbox that gives you code execution, a live browser, an AI agent, and human handoff in one place. Connect through the SDK and keep the browser in the same workflow context.

~1.7s
Warm claim
AI
Agent
Handoff
Human-in-loop
example.ts
1import { BanataSandbox } from "@banata-boxes/sdk"
2
3const sandbox = new BanataSandbox()
4const session = await sandbox.launch({
5 capabilities: {
6 agent: { enabled: true },
7 browser: { viewport: { width: 1280, height: 800 } },
8 },
9})
10
11const preview = await session.getPreview()
12const await session.promptAsync(
13 "Open example.com and report the title."
14)
Browser automation

Built for serious
web automation

The browser lives inside the same sandbox as your files, commands, and AI agent, so web work and workspace work stay connected.

Region-aware placement

The live browser launches inside the sandbox and stays connected to the same workflow context. Preferred region placement helps keep work close to where you need it.

Pass a preferred region during sandbox creation
See the actual assigned region once the sandbox is ready
Browser and workspace stay in the same session context
Works with warm reserve and resumed sandboxes

Warm reserve access

Prepared warm sandboxes reduce wait time for the live browser and workspace when work begins.

Warm pool maintained automatically
Fast claim path for ready capacity
Pause and resume supported on the same sandbox
Built for bursty real-world workflows

Controlled browser sessions

Run browser work inside the sandbox with controlled handoff, recording, and optional bring-your-own proxy on supported plans.

Team plan supports low-resolution recording artifacts
Team plan supports bring-your-own proxy
Control can move between AI and human
Browser state persists with the sandbox session

Live browser viewing

Watch the browser in real time, connect automation, or hand the screen to a human when the workflow needs manual input.

Live browser preview in the dashboard
Browser control can switch without leaving the page
Human handoff for login, approval, and ambiguous steps
Recording artifact support on the highest tier

Isolated workspaces

Every browser runs in its own micro-VM with a dedicated IP. No session can see another — by design.

Dedicated micro-VM per session
Environment variables scoped to the sandbox session
Checkpoints and artifacts available through the SDK
Pause, resume, and end without sharing state across users

Scale to thousands

Scale sandbox workflows with warm reserve, live preview, an AI agent, and artifacts on the same runtime substrate.

Tiered concurrency for sandbox workflows
Warm reserve maintained automatically
Active-hour pricing for predictable usage
Pause, resume, checkpoints, and artifacts built in
Code sandboxes

Secure execution with built-in tools

Spin up isolated workspaces to run code safely. Use Python, Node, Bun, shell tools, and files in the same sandbox that powers your browser workflow.

~ 50ms
Warm start
Built-in
Tooling
10GB
Storage
example.ts
1import { BanataSandbox } from "@banata-boxes/sdk"
2
3const sandbox = new BanataSandbox()
4const session = await sandbox.launch({
5 capabilities: {
6 browser: { viewport: { width: 1280, height: 800 } },
7 documents: { libreofficeHeadless: true },
8 },
9})
10
11const result = await session.exec(
12 `import pandas as pd
13 print(pd.read_csv("data.csv").head())`
14)
Code sandboxes

Run any code in
total isolation

Sandboxes are perfect for AI agents that generate code, data processing pipelines, or any workflow that needs secure server-side execution.

Built-in tooling

Use Node, Bun, Python, shell tools, Git, and common Linux utilities without choosing a runtime up front.

Languages and tools are already available in every sandbox
No runtime picker at sandbox creation time
Use exec, runCode, and terminal access from the SDK
One stable sandbox shape across plans

50ms warm starts

Pre-warmed sandbox pools mean your code starts executing almost instantly on the warm path.

Pre-warmed pools for the supported sandbox shape
Fast claim path when warm capacity is available
Session persistence for iterative execution
Usage is controlled by plan limits and active hours

Package installation

Install what you need with pip, npm, or bunx. Dependencies persist for the lifetime of the sandbox.

pip install, npm install, bunx — all supported
Persistent filesystem across executions
Works alongside files, documents, and browser tasks
Outputs can be saved as artifacts

File system access

Read, write, and manage files inside the sandbox. Upload input files, generate outputs, and download results.

Full POSIX filesystem per sandbox
Read, write, and list files through the SDK
Download generated files and artifacts
Document conversion outputs stay in the same workspace

Execution workflow

Stream stdout and stderr in real-time via WebSocket. Watch your code execute live — perfect for long-running tasks.

Command execution and code execution through the SDK
Browser work and workspace files stay in sync
AI agent tasks can drive both browser and workspace steps
Exit code and error capture

Composable with browsers

Use the live browser and the sandbox in one workflow. Browse, extract, process, and return results without leaving the session.

Use one sandbox with a paired browser
Pass data between the browser and workspace
Orchestrate multi-step workflows via SDK
Shared session context for seamless handoff
Use cases

Built for teams that
automate everything

Whether you're building AI agents, scraping pipelines, or testing frameworks, Banata gives you one runtime with code execution, paired browser control, and human handoff.

AI agent tooling

Give your AI agents real browsers and code execution. Let them browse the web, fill forms, extract data, and run analysis code — all through tool calls.

LLM tool useBrowser actionsCode execution

Web scraping at scale

Scrape any website with stealth browsers and residential proxies. Process extracted data in sandboxes. Handle millions of pages per day.

Anti-detectProxy rotationData pipelines

End-to-end testing

Run your Playwright or Puppeteer test suites in the cloud. Parallel execution across regions. Full session recordings for debugging failures.

Parallel testsCross-regionVideo recording

Secure code execution

Run user-submitted or AI-generated code safely. Each execution is fully isolated with resource limits, network controls, and automatic cleanup.

User codeAI outputResource limits
Technical specs

Numbers that
speak for themselves

Built on bare metal infrastructure optimized for browser and sandbox workloads. Every millisecond matters.

Region-aware
Placement
Preferred region support
Warm
Browser reserve
Prepared for fast claims
~ 50ms
Sandbox warm start
Pre-warmed pools
25
Concurrent sandboxes
Team plan
Live
Human handoff
In-browser control
Files
Artifacts
Saved outputs and downloads
Built-in
Tooling
Node, Bun, Python included
Plan-based
Usage limits
Concurrency + active hours

Ready to start building?

Start with Starter at $25/month for 3 concurrent sandboxes and 30 active hours.