
Plannotator
Plan and code review for your coding agents
About Plannotator
Plannotator is a local review tool for coding agents, published by backnotprop and licensed MIT or Apache 2.0 at your option. It puts a person back in the loop at the two moments that matter: it opens the agent's plan so you can annotate it before anything executes, and gives you a pull-request-style diff viewer over the code the agent wrote before you commit it. Your annotations go back to the agent as structured feedback. It installs as a CLI with hooks and slash commands for Claude Code, Codex, Copilot, Gemini, OpenCode, Pi, Amp, Droid, Kiro and VS Code.
You install the CLI with a single curl command and add the plugin to your agent. When the agent leaves plan mode, plan review opens by itself; you select text and mark it for deletion, add an inline comment, or write a replacement, and the annotations export as structured feedback the agent understands. For code, a review command opens a side-by-side or unified diff of uncommitted changes or a GitHub or GitLab pull request, where you annotate individual lines and stage files before committing. Review data stays on your machine by default.
- •Plan Review Before Execution - Hooks the agent's plan step so you can comment, mark deletions or write a replacement before any code is written.
- •Agent Diff Viewer - Side-by-side or unified diffs of uncommitted changes or a pull request URL, with line-level annotations and code suggestions.
- •Plannotator Inbox - One local window collecting the questions, plans and finished work your agents leave, so no agent session waits on your reply.
- •Runs Locally - Review data stays on your machine by default, with no usage telemetry or analytics, and encrypted short links only when you share.
- •Editor and Notes Integrations - Opens plans in VS Code tabs and saves them to Obsidian or Bear with frontmatter, tags and project metadata.
Developers who run coding agents on real codebases and want a say before the agent acts, not only after. It suits teams that already review human pull requests and want the same discipline applied to agent output. It is less useful if your agent exposes no plan step to hook, or if you would rather let the agent run unsupervised.
Pricing
No pricing page found
- Free and open source$0/mo





















































