OpenCode Browse connects AI agents to a person’s existing browser through an MCP server and Chromium extension. Agents can use the browser’s active logins and installed extensions, avoiding separate credentials or account setup. They open tabs in the background and keep them grouped out of the way, then read pages, click controls, enter text, take screenshots, record tabs, and capture network traffic. Supported browsers include Chrome, Edge, Brave, Arc, Helium, Vivaldi, Opera, and Chromium; the tool can be used with MCP clients such as OpenCode, Claude Code, Cursor, Codex, and Zed.
The design includes ways to keep people in control. Users can restrict agents to named sites, lend an agent an already-open tab with the extension’s toolbar button, or take over when a task requires a login, two-factor authentication, or payment. After completing the step, they click Continue so the agent can resume. Setup starts by installing the Chrome extension, which guides users through the remaining steps. A command-line installation is also available through shell scripts or package managers such as npm and Bun, though the extension-first route is recommended.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.