BROWSER AGENT

Explore The Features

Give a goal. It plans, acts and delivers.

Available on all plans. See pricing for limits.

ASK & ACT

Ask. Plan. Act. Deliver.

Start from Claude Code, Codex, Cursor or the built-in Chat. It plans, works in the DsphereAI browser and hands back the result.

Ask Plan & Act Deliver

Give A Goal

Describe the job in one sentence, in your agent or the built-in Chat.

Side panel Chat with a collection request typed beside the Books to Scrape page, attached with @

Plan, Then Act

It shows a plan, then works in the DsphereAI browser, reusing the sessions you signed in to there. Sensitive actions wait for you to confirm, and you can pause or stop anytime.

Live run progress: finished steps checked off, the current step running

Deliver Results

Data goes to local data sheets or export files, and results go back to your agent or Chat.

Collected results: 200 books with price, stock status and title

RUN & SCHEDULE

Run In Parallel

Independent runs. Isolated sessions.

History panel with three demo tasks running at the same time

Isolated Sessions

Parallel Runs

Task Groups

Live Progress

SCHEDULE & REVIEW

Schedule Runs

Run on a schedule or on a set date.

Run guide →

WATCH & APPROVE

Watch Every Step. Approve What Matters.

See the plan, every action and the page it acted on. Sensitive steps wait for your OK, and you can pause or stop anytime.

Every action is checked

A page frame for each step

You confirm what matters

Live viewStep logConfirmPauseStop

A running task in DSphere: verified steps with a page frame for each one

Confirmation sheet with Confirm and Don't Do It

SKILLS & AGENTS

Built-In Skills, Any Agent

01 CHAT SKILLS

Ready-Made Skills

Scrape data, fill forms, download files. Pick one from the + menu.

See it in action

02 AI CONNECTORS

Connect Your Agent

Give the Skill setup prompt to Cursor, Claude Code, Codex or WorkBuddy to install and configure MCP with the guide, or use the built-in Chat.

Claude

Codex

Cursor

Skill → MCP

View connectors

SECURITY & CONTROLS

Isolate And Control

Manage sessions, proxies and fingerprints.

Two isolated sessions, each running in its own tab group

Private Sessions

Run in an assigned profile.

Proxy Routing

Set a proxy per session.

Fingerprints

Reuse identity settings.

Isolated Sessions

Separate cookies, storage and state.

Browser Execution

Navigate, interact and download.

Authorized Extraction

Follow site rules and rate limits.

Credentials & MFA

Stored separately, used by reference.

Run Records

Spot issues in status and logs.

Use authorized pages and data only.

STEALTH MODE

Your Fingerprint. Your IP.

Give every account its own browser fingerprint and proxy IP, so each one looks like a separate, real device.

Browser fingerprint settings: engine, operating system, user agent and an overview of fingerprint signals — Browser Fingerprint
Choose the engine, OS and user agent, then tune 20+ signals such as WebRTC, Canvas, WebGL, fonts and time zone.

Proxy IP settings: IP channel, protocol, host and port, proxy account and linked application accounts — Proxy IP
Bind a proxy such as SOCKS5 to each account, pick an IP channel and test it before use.

COMPARISON

Browser Agent vs. RPA

Capability

Traditional RPA

Browser Agent

Task creation

Studios, selectors and scripts

✓Plain-language goals via your agent or Chat

Runtime environment

Desktops, VMs and runners

✓Assigned browser profiles and tab groups

Parallel scale

Robot licenses and runner capacity

✓Parallel workflows in isolated tab groups

Web data

UI scraping, OCR or external tools

✓Structured extraction, local sheets and AI

Monitoring and debugging

Orchestrator consoles and robot logs

✓Live task card: plan, page frames, Pause / Stop

Use both: Browser Agent for web tasks, RPA for desktop workflows.

RUN YOUR FIRST TASK

Put The Features To Work

Give a goal in your agent or the built-in Chat, and let it plan, act and deliver your first browser task.

Download for free Explore use cases Compare plans and features

FAQ

Can tasks run on schedules or in parallel?

Yes. Save a verified task, schedule recurring or date-specific runs, and run tasks in parallel. See pricing for limits.

What if a login expires, a CAPTCHA appears or a page changes?

It reuses signed-in sessions and the 2FA authenticator, so you don’t have to sign in again. Model Repair handles page changes, and sensitive actions ask you to confirm.

Where can extracted data go?

Local data sheets, CSV or Excel export, downloaded files and images, or email. Results also return to your agent or Chat.

How do I watch and control a run?

A live task card shows the plan and page frames, with Pause and Stop. Run history and logs stay available.

Does a local data sheet mean the AI runs locally?

Not necessarily. Model location and data scope depend on your connection settings. Follow your team’s data policies.