Give opencode eyes: screenshot any URL or document with pixelshot and read it visually.
Adds a screenshot tool that renders a page to image tiles (Playwright/CDP) so the agent
sees charts, diagrams, tables, and layout the way a person does.
pip install pixelrag # provides the pixelshot command (or: uv tool install pixelrag)Then add the plugin to your opencode.json:
{
"$schema": "https://opencode.ai/config.json",
"plugin": ["@startrail/pixelbrowse"]
}opencode installs npm plugins automatically at startup. For local development from a
clone, symlink or copy index.js into ~/.config/opencode/plugins/ (global) or
.opencode/plugins/ (project) — files there are loaded directly at startup.
Just ask opencode to look at a page:
opencode run "screenshot https://news.ycombinator.com and summarize the top stories"
The agent calls the screenshot tool with a URL (or a local HTML file, PDF, or image),
gets back the tile image paths, and reads them visually. Optional tool args: output
(tile directory, default .opencode/pixelbrowse inside the project) and viewportWidth (default 875; use 1280
for desktop layouts).
No MCP server, no backend — the plugin just calls pixelshot on your machine.
The package is published manually from this directory:
npm publish --access public