DOCS

How pinlet works

A browser extension pins comments to elements on your site. A dashboard collects them. Coding agents read and answer them over MCP.

Install the extensionAdd it to Chrome, sign in, add your site. Four steps. Agents and MCPPoint Claude Code or Codex at your project and let it work the inbox. Self hostingPostgres and a Node server. Docker compose for the database.

The short version

Projects

A project is your site, described by one or more address patterns: production, previews, and the developer's own machine. The extension only wakes up on addresses that belong to a project you are in. Everywhere else it does nothing.

Threads and pins

Press c, click an element, type. pinlet stores a fingerprint of that element, not a screenshot of where it was, so the same pin shows up on every address of the project, even when the data on screen is different. A cropped screenshot comes along for context.

Technical details

The composer has one checkbox, Attach technical details. Ticked, the last two minutes of console messages and request summaries are stored with the thread, which is tagged bug. Never request bodies, never headers. Unticked, nothing leaves the page.

The inbox

The dashboard at pinlet.dev/app lists every thread, grouped by page, with filters for status, environment and the bug tag. Each thread links back to the page with the pin open.

Agents

Each project can issue API keys. A coding agent uses one to list open threads, read the evidence, make the change and reply. Agent use is never metered. Set it up.