A Pastel alternative built around your coding agent.
Pastel collects comments on a shareable copy of your site so a human can act on them later. Pincushion pins live on the deployed app itself — and each pin becomes an agent work packet. Selector, DOM snippet, screenshot, viewport, thread: one MCP call, and Cursor, Claude Code, or Codex ships the change from your IDE.
What Pastel does well
Pastel's pitch is simplicity: paste a URL, get a share link, and reviewers comment on the page (or a PDF or image) with zero installs and no accounts. For agencies collecting copy edits and design notes from clients, that link-based flow is genuinely low-friction, and Pastel exports comments to Trello, Asana, Jira, and similar tools. The trade-offs come from the same design: reviewers comment on Pastel's rendered copy of your page rather than the live app itself, the free plan limits how long commenting stays open (a few days at the time of writing), and paid plans are priced per user — starting around $24/month for a single seat, more for studio tiers. Check their pricing page for current numbers.
Where the model differs
Pastel's output is a list of comments a human reads, interprets, and re-types into tickets or into a prompt for their coding agent. That translation step is exactly what Pincushion deletes. A Pincushion pin isn't a note about the page — it's implementation context: the CSS selector of the pinned element, the surrounding DOM, a screenshot, the viewport it was seen at, and the full discussion thread. Your agent reads all of it via MCP with implement_approved_pins, makes the change, and records the branch, commit, and PR on the pin. The feedback loop ends in a deploy, not an export.
There's a second, subtler difference: because pins anchor to elements on the live app, they survive reloads, dynamic state, and responsive layouts. Feedback on a static snapshot goes stale the moment you ship; feedback anchored to a selector on production doesn't. More on that in our guide on why canvas-anchored comments break on deployed apps.
How Pincushion differs
| Pincushion | Pastel | |
|---|---|---|
| Where feedback lives | On the deployed app itself (extension or script tag) | On a shareable rendered copy of the page |
| Pin payload designed as agent work packet (selector + DOM + screenshot + viewport + thread) | Yes | Comment + position |
| Coding-agent integration (Cursor, Claude Code, Codex via MCP) | First-class | No — exports to PM tools |
| PDFs, images, video review | No — live web apps only | Yes |
| Reviewer seats | Free, unlimited, forever | Reviewers comment free; makers pay per user |
| Branch / PR / deploy trail on resolved feedback | Yes | No |
| Free tier | Unlimited manual pins, 1 project, 25 AI actions/mo | Free plan with a short commenting window |
| Starting paid price | $19/mo Pro (14-day trial on signup) | Per-user plans from about $24/mo |
Pastel gets comments out of email threads. Pincushion gets them into your coding agent — with the selector, DOM, and screenshot already attached.
When Pastel is the better fit
If your review workflow covers more than live web apps — PDFs, image mockups, video — Pastel handles those and Pincushion doesn't. And if the people implementing feedback aren't using AI coding agents at all, Pincushion's core advantage (the agent work packet) matters less to you. Pastel is a good comment-collection tool. The gap only opens when the fix is going to be made by Cursor, Claude Code, or Codex — then a comment on a snapshot is two translation steps away from done, and a Pincushion pin is zero.
Feedback that ends in a PR, not an export
Reviewers are free and unlimited. The free developer tier covers one project with unlimited manual pins and 25 AI actions a month, and every signup starts with a 14-day Pro trial: unlimited automation, realtime push to your IDE, unlimited projects, and access controls. See it working in Cursor, Claude Code, or Windsurf, or compare with Markup.io and ruttl.