OpenCode launches Browse, a browser automation tool for AI agents
OpenCode has released Browse, consisting of an MCP server and a Chromium extension that let AI agents operate a user's actual browser session, including existing tabs, logins and installed extensions. Agents can open background tabs, read page content, click, type, take screenshots and record activity, then hand control back to the human for steps like login, two-factor authentication or payment before resuming once the user clicks Continue.
GoKawiil's interpretation of the reporting above, not reported fact.
By using the real logged-in browser rather than a sandboxed instance, Browse could let agents complete tasks that require existing sessions or cookies, such as checking email or using subscription services, without re-authenticating. The handoff design for sensitive steps suggests an attempt to keep humans in the loop for security-critical actions, though how robust that separation is in practice remains to be seen.
- Browse pairs an MCP server with a Chromium extension to give agents browser access.
- Agents can read, click, type, screenshot and record tabs opened in the background.
- Control returns to the user for logins, 2FA and payments before the agent resumes.
Source: opencode.ai, 2026-11-16
Published there as: “OpenCode Browse”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.