5 items
Annotate any page and send the element, screenshot, and your comment to your local coding agent Point at what's wrong. Your coding agent gets the details. Click an element in the page, type what should change, and hyzer-annotate sends your comment, the element's selector and metadata, the page URL, and a screenshot cropped to the thing you pointed at — straight to a coding agent running on your own machine. No more describing a layout bug in prose, or pasting a full-page screenshot and hoping the agent finds the button you meant. HOW IT WORKS hyzer-annotate talks to a small MCP server you run locally, alongside your agent. The extension finds it on localhost, you connect a tab, and everything you annotate is queued for the agent to pick up when you ask. Works with any coding agent that speaks MCP. For example, with Claude Code, install the server with: ``` claude mcp add hyzer-annotate --scope user -- npx -y @hyzer-labs/annotate ``` WHAT YOU GET • Element mode — click any element, comment on it, and the screenshot is cropped to that element with a little context around it. • Page mode — comment on the page as a whole, with the full scrollable page stitched into one screenshot. • A running count of what's queued, so you know what the agent hasn't seen yet. • Optional persistence, so a queue survives restarting your agent. • Screenshots are optional per annotation, when the comment says it all. NOTHING LEAVES YOUR MACHINE The only network destination in the entire extension is 127.0.0.1 — the server you started yourself. There is no account, no telemetry, no analytics, and no remote endpoint. Site access is requested one origin at a time, when you connect a tab, never at install. OPEN SOURCE GPL-3.0. Source, issues, and the MCP server: https://github.com/hyzerlabs/annotate
rating_count is the Chrome Web Store ratings count, not a written-review count.