Plannotator

About Plannotator

Plannotator is a locally-run review surface for AI coding agents, addressing the two moments where agent work usually goes wrong: approving a plan you skimmed, and merging a diff you never really read. On the plan side it hooks into the agent's plan step automatically where hooks are supported, and offers slash commands where they are not. Instead of squinting at terminal text, you get a proper workspace: select any passage, mark it for deletion to cut scope, leave an inline comment, or write a replacement. Your annotations export as structured feedback that goes straight back to the agent — no copy-pasting or retyping — and every revision is saved with version history and a diff view showing what changed between iterations. On the code side, /plannotator-review opens a PR-style viewer over your agent's uncommitted local changes or a GitHub/GitLab PR URL, with side-by-side or unified diffs, file-tree navigation, line-level annotations that can carry code suggestions, and staging or unstaging before you commit. It also annotates arbitrary markdown, folders, or URLs, and the agent's last message. Everything runs on your machine, so plans never leave it; shared links are encrypted with the data living in the link itself. There are integrations for VS Code (editor tabs, diff view, editor annotations), Obsidian (auto-save to your vault with frontmatter and tags), Bear, and GitHub PRs, and it works with Claude Code, Codex, Copilot, Amp, Droid, Gemini, Kiro, OpenCode, and Pi. The local plugin is free and open source under MIT/Apache-2.0; a hosted team layer called Workspaces is on a waitlist.

Key Features

Plannotator Pros & Cons

Who Is Plannotator Best For?

Tags