Repository navigation
Plannotator
AI coding agents propose a plan and then change code, and the human often approves both without a structured way to push back. Plannotator is a review tool that puts a proper review step in front of the agent's plan and its resulting code changes, and returns the reviewer's comments to the agent in a form it can act on.
For plan review, the tool hooks into the moment an agent finishes planning. When an agent calls ExitPlanMode, a permission hook starts a local server that reads the plan and opens a review page in the browser. The reviewer annotates the plan and approves or denies it. A denial sends structured feedback back so the agent can revise, and a diff shows what changed between versions (README).
For code review, it covers local changes and GitHub or GitLab pull and merge requests, and it can also annotate specs, markdown, HTML, and URLs. The product site describes inline comments, deletion marks, replacement text, a side-by-side or unified diff viewer, line-level annotations, and draft auto-save.
The README lists support for Claude Code, Codex, Copilot CLI, Gemini CLI, OpenCode, Kiro, Droid, Amp, and Pi, with integrations for VS Code, Obsidian, and Bear, plus remote and SSH use. It installs through a single script that sets up hooks, skills, and slash commands for the agents it detects. The README states that it collects no telemetry, although it checks GitHub for new releases when pages load.
The repository was created on 2025-12-28 and carries the Apache-2.0 license. At the time of review it had 9469 stars and 165 open issues.
Plannotator is placed in Assess within Tools. It addresses a concrete gap in agent workflows, the lack of structured human review before and after an agent acts, and its adoption signals are strong for a project under a year old. No first-person production use has been confirmed, so the ring remains Assess.
