plannotator: A browser-based review layer for AI agent workflows

A local, browser-based review tool that connects directly to various AI coding agents such as Claude Code, Codex, and Copilot CLI. When an agent generates plans, specs, HTML artifacts, or code changes, they automatically open in the browser, allowing users to add markup and annotations to send feedback. Sent feedback is immediately passed to the next turn of the corresponding agent session, continuing the revision process.

Plannotator Inbox serves as an asynchronous review channel between the agent and the user. When an agent asks a question or leaves a file, you can review it in the inbox and send a response with a single click, maintaining the workflow without keeping the session open. It integrates with various version control systems, including Git, GitButler, Jujutsu, Perforce, GitHub, and GitLab, supporting code reviews from local changes to remote PRs.

All data is stored locally by default, with no telemetry collection. Network communication occurs only when explicit features are used, such as checking for the latest release, URL annotation, AI questions, or generating share links. The sharing feature applies AES-256-GCM encryption to protect sensitive content, and team collaboration is extended through hosted Workspaces.

It is suitable for developers who use AI agents to write code or create plans. It is useful when you want a middle stage where a human reviews and guides the direction, rather than simply accepting the agent's output. It provides the same review experience in the terminal via Herdr Annotate, integrating the records of app reviews and terminal reviews.