The /plannotator-annotate command opens any markdown file in the Plannotator annotation UI.

Usage

Slash command (inside an agent session)

/plannotator-annotate path/to/file.md

The agent runs plannotator annotate <file> under the hood. The annotation UI opens in the browser. When you submit, feedback is returned to the agent as structured output.

Standalone CLI (outside an agent session)

plannotator annotate path/to/file.md

This starts a local server, opens the browser, and blocks until you submit. The formatted feedback is printed to stdout.

How it works

User runs /plannotator-annotate README.md
        ↓
CLI reads README.md from disk
        ↓
Annotate server starts (random port)
        ↓
Browser opens, loads annotation UI
        ↓
/api/plan returns { plan: markdown, mode: "annotate" }
        ↓
User annotates → Send Annotations
        ↓
POST /api/feedback with exported feedback
        ↓
Server prints feedback to stdout, exits

Annotate mode differences

The annotation UI in annotate mode works the same as plan review, with a few changes:

• The “Approve” button is hidden (there’s nothing to approve)
• “Send Feedback” becomes “Send Annotations”
• Cmd/Ctrl+Enter sends annotations instead of approving
• The completion screen says “Annotations Sent” instead of “Plan Approved”

All annotation types work identically — deletions, replacements, comments, insertions, global comments, and image attachments.

Feedback format

When you send annotations, they’re exported as structured markdown:

# Plan Feedback

I've reviewed this plan and have 2 pieces of feedback:

## 1. Remove this
` ` `
the selected text
` ` `
> I don't want this in the plan.

## 2. Feedback on: "some highlighted text"
> This needs more detail about error handling.

---

The agent receives this and can act on each annotation.

Server API

Environment variables

The annotate server respects the same environment variables as plan review. See the environment variables reference.