OpenCode Open-Source Browser Control Tool: AI Takes Over Real Chrome, Adds Anti-Bot Detection Design

Beating AI News Flash: OpenCode has officially launched Browse, enabling AI Agents to directly operate browsers such as Chrome that users use daily. It connects tools like Codex, Claude Code, and OpenCode through MCP and browser extensions, allowing AI to reuse existing tabs, login states, and extensions without launching a separate browser for AI.

Agents can open their own tabs in the background, using Playwright to click, fill out forms, take screenshots, record screens, and log network requests without affecting the user's normal browsing. When encountering steps that require the user's personal action, such as login, two-factor authentication, or payment, control can be handed back to the user.

Browse also supports cross-device control. For example, an Agent running on a remote server can operate Chrome on the user's MacBook that is already logged into an account, without needing to migrate passwords or the entire browser session to the cloud. Cross-device connections require additional encryption tools such as Tailscale.

The OpenCode developer has also specifically added anti-detection mechanisms. The project uses Patchright to reduce some automation features exposed by the browser, and by default simulates the rhythm of human mouse movement, typing, and scrolling, attempting to reduce the probability of being identified as a bot by websites.

However, these measures cannot guarantee bypassing website risk controls. When encountering CAPTCHAs or anti-bot interception, Browse will identify the anomaly and hand over steps requiring manual verification to the user. The project has been open-sourced under the MIT license.