OpenCode, an open-source browser control tool, uses AI to control a real Chrome browser and incorporates anti-bot detection features.

Beating AI News: OpenCode officially launches Browse, enabling AI Agents to directly operate users’ daily browsers like Chrome. It connects tools including Codex, Claude Code, and OpenCode via MCP and browser extensions, eliminating the need to launch a separate browser for AI while reusing existing tabs, login sessions, and extensions. Agents can open their own tabs in the background, perform actions like clicking, form filling, screenshotting, screen recording, and network request logging via Playwright, without disrupting users’ regular browsing. For steps requiring manual user input, such as login, two-factor authentication, or payment, Browse can hand back control to the user. Browse also supports cross-device control: Agents running on remote servers can operate Chrome on a user’s MacBook (with existing logged-in accounts), without requiring password or full browser session migration to the cloud. Cross-device connections require additional encryption via tools like Tailscale. OpenCode’s developers have also added anti-detection mechanisms. The project uses Patchright to reduce visible automated browser features, and by default simulates human-like mouse movement, typing, and scrolling rhythms, to lower the chance of being flagged as a bot by websites. However, these measures do not guarantee evading website risk controls. When encountering CAPTCHAs or anti-bot blocks, Browse detects anomalies and hands off manual verification steps to users. The project is open-sourced under the MIT License.