SeggWat LogoSeggWat
BlogDashboard

Review Mode

Close the loop between what your AI built and what you actually wanted — file review tasks, drop on-page pins, and read them back from the CLI, MCP, or dashboard.

Overview

Review Mode is an on-page review loop for AI-assisted development. It answers one question: did we build the right thing?

The loop has three steps:

  1. A task is filed — you or your AI create a review task ("Check the new pricing page") from the dashboard, CLI, or MCP.
  2. A reviewer pins on the page — open the running app with the review widget active, and drop numbered pins (with comments and freehand drawings) directly on the UI that's wrong.
  3. The pins are read back — your AI (or you) reads the annotations from the CLI, MCP, or dashboard, fixes each one, and marks it resolved.

Every pin is anchored to a real DOM element and stored with its page path, so a comment like "this button is too faint" points at exactly the button.

Three ways to run it

Hosted

The widget runs on your real (staging/preview) site against a SeggWat project. Pins land in your dashboard, CLI, and MCP.

Local proxy

seggwat review serve — annotate a localhost dev app against a real project without touching allowed-origins.

Standalone

seggwat review local — a fully offline loop backed by ./.review/ JSON. No account, no project, no API key.

Pick based on whether you have a SeggWat account and where your app runs. All three drive the same widget and the same pin format.

Hosted

Use this when your app is reachable at an origin you can add to the project (a staging URL, a preview deploy, or production).

1

Add the review widget to your app

html
<script src="https://seggwat.com/static/widgets/v1/seggwat-review.js"
            data-project-key="YOUR-PROJECT-KEY"></script>

The overlay stays dormant until a review task is in the URL, so it's safe to leave in staging/preview builds. Add the page's origin to Settings → Project → Allowed origins so the widget can talk to the API.

2

Create a review task

From the dashboard's Review page, the CLI, or MCP:

bash
seggwat review create PROJECT_ID -t "Review the new pricing page" \
      -d "Check the hero, the plan cards, and the CTA"

Copy the returned task id.

3

Open the page in review mode

Append ?seggwat-review=<task_id> to any URL on your site:

https://staging.yourapp.com/pricing?seggwat-review=6a4faf9b3a8d3cd58051203b

A toolbar appears. Pick the pin tool, click the element that's wrong, type a comment, and optionally draw an arrow or box. Repeat for every issue.

4

Read the pins back and fix them

bash
seggwat review comments PROJECT_ID TASK_ID

Each pin shows its number, the pinned element's selector, the page path, and the comment. Fix an issue, then mark it resolved:

bash
seggwat review resolve PROJECT_ID FEEDBACK_ID

Local dev proxy

Your dev app runs on localhost, but the review widget's endpoints are origin-gated — a localhost origin isn't on your project's allow-list. seggwat review serve solves this without any config change: it runs a tiny localhost proxy that forwards the widget's calls to SeggWat with a trusted Origin, so pins land on your real project.

bash
seggwat review serve PROJECT_ID          # defaults to port 7777

It prints a ready-to-paste snippet — add it to your dev app, then open a page with ?seggwat-review=<task_id> as above:

html
<script src="https://seggwat.com/static/widgets/v1/seggwat-review.js"
        data-project-key="YOUR-PROJECT-KEY"
        data-api-url="http://localhost:7777"></script>

The proxy uses your project's first allowed origin by default; pass --origin https://yourapp.com to choose a specific one. Read pins back with the same seggwat review comments command.

Standalone (no account)

Want the review loop with no SeggWat account at all — just you, your AI, and your local app? Use seggwat review local. Pins are stored as plain JSON in ./.review/ in the current directory. There's no project id and no API key.

1

Start the local server

bash
seggwat review local serve          # widget server on :7777
2

Create a task and add the snippet

bash
seggwat review local create -t "Review the checkout flow"
html
<script src="https://seggwat.com/static/widgets/v1/seggwat-review.js"
            data-project-key="local"
            data-api-url="http://localhost:7777"></script>
3

Pin, then read back

Open your app with ?seggwat-review=<task_id> and drop pins. Then:

bash
seggwat review local comments TASK_ID
    seggwat review local resolve  FEEDBACK_ID

Reading pins from anywhere

However pins are created, you can read and resolve them from any surface:

Surface Create a task Read pins Resolve
Dashboard Review page Expand a task to see its pins Resolve/reopen buttons
CLI review create review comments review resolve / reopen
MCP create_review_task list_review_task_comments update_feedback → status: Resolved

See the CLI guide for every review and review local command, and the MCP guide for the review-task tools your AI assistant can call directly.

Navigation