Desktop agent

An AI agent that works directly in your browser

It understands tasks, works on web pages, builds automation, and connects local tools and team services inside authorized profiles.

From browser actions to finished results

These capabilities use the app's browser tools, automation, local commands, skills, and MCP connections.

Open the right project and session, then navigate, click, type, select, scroll, and drag on the page.

Move across authorized store profiles and handle tabs, popups, waits, and multi-page tasks.

Read page content, capture screenshots, upload files, and verify results.

Create or update automation from a task goal, select profiles, and run a guided test.

Run approved local commands to process CSV, JSON, images, and result files.

Install local skills to give the agent your team's knowledge, rules, and standard processes.

Connect team Streamable HTTP services with the required credentials and access tokens.

Enable the agent by team and configure model providers, headless mode, and command approvals.

From instructions to verification, the agent selects the tools it needs within its authorized scope.

Task brief

Describe the stores, pages, actions, and expected output in plain language.

Smart execution

It enters authorized profiles, reads page state, and completes actions and data processing step by step.

Review results

Review the agent's data, screenshots, documents, and errors at any time.

The agent operates the product; it does not replace fingerprint isolation, proxy management, or team access controls.

Shell Browser

Desktop agent

Download the desktop app and use the agent in an authorized project profile.