Plannotator: Visual Plan & Code Review for AI Coding Agents

⬅️ Back to Tools

Plannotator: Visual Plan & Code Review for AI Agents

Plannotator (4.9k+ stars) is an open-source tool that lets you visually annotate and review AI coding agent plans and code diffs. No more guessing what your agent will build, mark up plans, leave feedback, and send structured notes directly back to the agent.

Works with six platforms: Claude Code, Copilot CLI, Gemini CLI, OpenCode, Pi, and Codex.

Why This Exists

AI agents write code fast but often miss edge cases or skip spec requirements. Plannotator gives you a visual UI to mark up plans and diffs before the agent implements anything. Your feedback goes straight back to the agent as structured input.

Key Features

Visual Plan Review

When your agent enters plan mode, Plannotator opens a browser UI where you delete, insert, replace, or comment on any part of the plan. Approve to proceed, or send annotations back for revisions.

Automatic Plan Diffs

See exactly what changed when the agent revises a plan. No manual diffing of markdown files.

Code Review Mode

Use /plannotator-review to view git diffs or GitHub PRs. Comment on specific lines of code, run built-in AI reviews, and send feedback to the agent.

Annotate Anything

  • /plannotator-annotate <file|folder|url>: Mark up markdown files, HTML, URLs, or entire folders
  • /plannotator-last: Annotate the agent’s last response and send feedback immediately

Team Sharing

Share plans and annotations with colleagues via private links. Small plans encode entirely in the URL hash (zero server storage). Large plans use end-to-end encrypted pastes (AES-256-GCM) that auto-delete after 7 days. No server can read your data.

Supported Platforms

PlatformInstall Method
Claude CodeInstall plannotator CLI, then /plugin marketplace add backnotprop/plannotator
Copilot CLIInstall plannotator CLI, then /plugin marketplace add backnotprop/plannotator
Gemini CLIInstaller auto-detects Gemini, enables plan hooks automatically
OpenCodeAdd @plannotator/opencode@latest to opencode.json, run install script
Pipi install npm:@plannotator/pi-extension
CodexInstall plannotator CLI, hooks enable automatically on macOS/Linux/WSL

Installation

Quick Install (macOS / Linux / WSL)

curl -fsSL https://plannotator.ai/install.sh | bash

Quick Install (Windows PowerShell)

irm https://plannotator.ai/install.ps1 | iex

Platform-Specific Setup

After installing the plannotator CLI, follow the platform-specific steps:

Claude Code

/plugin marketplace add backnotprop/plannotator

Restart Claude Code after install.

Copilot CLI

/plugin marketplace add backnotprop/plannotator
/plugin install plannotator-copilot@plannotator

Restart Copilot CLI after install.

Gemini CLI

Plan review activates automatically when you use /plan. Requires Gemini CLI 0.36.0+.

OpenCode

Add to opencode.json:

{
  "plugin": ["@plannotator/opencode@latest"]
}

Run the install script, then restart OpenCode.

Pi

pi install npm:@plannotator/pi-extension

Start Pi with --plan to enter plan mode.

Codex

Works automatically on macOS/Linux/WSL. Use !plannotator review or !plannotator annotate in Codex Desktop.

How It Works

  1. Your agent finishes a plan or code diff
  2. Plannotator opens a browser UI automatically
  3. You annotate, comment, delete, or replace sections
  4. Approve → Agent proceeds with implementation
  5. Request changes → Annotations are sent back as structured feedback

Tradeoffs and Risks

  • Browser required: Review happens in a web UI, not your terminal
  • Early-stage for some platforms: Pi and Codex support is newer than Claude Code/Copilot
  • Windows hooks limited: Codex hooks are disabled on Windows, use direct commands instead
  • Self-hosting required for long-term sharing: Default paste service deletes after 7 days

Who Should Use It

Use Plannotator if you:

  • run AI coding agents and want to review plans before implementation
  • work in a team and need to share agent plans with colleagues
  • want structured feedback loops instead of freeform chat
  • use multiple agent platforms and want a consistent review workflow

Why This Tool Rocks

  • Multi-Platform: Works with 6+ major AI coding agents
  • Visual First: No more reading raw markdown plans in your terminal
  • E2E Encrypted Sharing: Team collaboration without trusting a server
  • Code Review Built In: Review git diffs and PRs, not just plans
  • Zero Server Storage: Small plans never leave your browser
  • Self-Hostable: Run your own paste service for long-term sharing
  • Open Source: Dual-licensed Apache 2.0/MIT, 4.9k stars, active development
  • Battle-Tested: Used by teams running production agent workflows

🔗 GitHub: github.com/backnotprop/plannotator

🔗 Website: plannotator.ai

Crepi il lupo! 🐺