Pinpoint
Visual review for AI agents.
Point at what is wrong in any UI and turn that visual feedback into actionable agent work.
What it is
Pinpoint is a visual review workflow for AI coding agents. You drop a pin on the exact part of an interface that feels wrong, add a short note, and hand the resulting review file to an agent.
A review keeps visual feedback, its exact location, and the agent-facing context together.
Who it is for
Pinpoint is for people shipping product UI with AI coding agents: engineers, designers, and product-minded reviewers who need feedback to stay precise when the person making the change is not sitting in the review.
Why I built it
Text-only UI feedback is too lossy. A comment like "the spacing feels off near the card header" still leaves the agent guessing which edge, which viewport, and which state matters. Pinpoint keeps the feedback attached to the surface that caused it.
The constraint
The workflow had to make feedback more actionable without turning a visual review into another project-management tool. It also needed to work across repositories and handoffs, where a reviewer may not share the agent's local context.
Product decisions
- Make the pin the unit of work. Each annotation should be small enough for an agent to fix and verify.
- Keep the artifact portable. The review should survive across tools, repos, and handoffs.
- Avoid turning review into a project-management system. The value is precision, not another workflow.
- Treat a review as an agent input, not a dead-end comment thread. The output should carry enough context for the next person or agent to act.
Technical notes
The useful part is not just storing coordinates. The tool needs enough context to let an agent understand the target, replay the feedback, and work through annotations one at a time. I built the product, interaction model, and MCP-facing workflow around that artifact boundary.
Status
Pinpoint is public and actively useful as a companion workflow for visual QA, design review, and agent handoff.