OpenCode Browser Plugin

OpenCode Browser connects your coding agent to a browser profile you choose. It works with OpenCode, Codex, Claude and other MCP clients through the opencode-chromium package.

Use it to: • Search and inspect pages without copying everything into a chat. • Navigate tabs, fill forms, use the keyboard and take screenshots. • Keep separate agent sessions organized in named browser profiles. • Review upload requests before files are sent to a website. • Save successful actions in optional local memory for repeat tasks.

The panel brings connection status, decision assistance and memory into three views: Overview, Profiles and Settings. Detailed usage and diagnostics expand when you need them. Light and dark themes follow your system.

Optional decision assistance can rank ambiguous targets with Jev through OpenRouter or TypeSafe, or OpenAI Luna through OpenRouter. Cloud assistance is off by default. Screenshot sharing needs a separate opt-in. Jev can use a small image as an experimental fallback after an unsure text decision. These services require your own API key and may charge for use.

Getting started:

1. Install the extension.
2. Run: npx -y opencode-chromium@latest setup --all
3. Restart your coding agent and choose the browser profile shown in the extension.

The extension requires the local native host installed by that command. It is a browser automation bridge, not a standalone chatbot. Requested page content and screenshots go to the agent you connect, which may send them to its model provider. Saved action memory stays on your device. Upload confirmation is on by default; turning it off allows uploads without a prompt. The extension has no telemetry.

Chromium browsers use the browser's debugger extension API. Firefox and LibreWolf also require the Remote Agent to be enabled at launch and configured on loopback. Setup instructions and troubleshooting: https://github.com/Quindart-com/opencode-chromium

Open source under the MIT license.