Plannotator opens a local diff viewer (git, jj, p4). You annotate lines, tokens, or comment on files, and when you're done they feed straight into your agent session; creating an automated feedback loop.

It works on local diffs, commits, branches, worktrees, and GitHub/GitLab PRs. The PR view lets you annotate the descriptions and comments while also being able to filter bot comments.

The core tool enables focused manual review, the optional AI layer:

  - Ask AI: Ask questions directly inline
  - Guided Reviews: Have an agent reorganize a large diff into semantic overview sorted by importance (inspired by linear)
  - Review Agents: Run review agents whose findings land as inline annotations. Review jobs can be your own skills that plannotator links to locally.

There's also a local API, your agents can POST findings or other markers/insights/notes onto the diff. Every annotation you make is saved to ~/.plannotator. Point an agent at that history and repeated feedback distills into a custom review skill (compound engineering).

Plannotator works with most agents and the AI features utilize core model subscription providers.

Free, MIT/Apache 2.0, runs entirely on your machine.

https://github.com/backnotprop/plannotator