Using PR Preview Deployments is the perfect way to interact with agentic coding output for visual inspection and providing feedback (along with screenshots!). Unfortunately the actual comment content and screenshots are locked in Vercel’s UI - completely inaccessible to CI automation.
The Opportunity
AI coding agents (Claude Code, Cursor, Copilot Workspace, etc.) are increasingly running in CI pipelines. These agents can read GitHub PR comments and act on them - but they can’t access Vercel preview comments.
This is a missed opportunity to augment and reward human inspection of previews. Right now, a human takes time to review a preview, spots an issue, leaves detailed feedback with a screenshot… and then has to wait for a developer to manually address it. The feedback loop is broken.
Proposed Solution
Either:
Auto-sync option: Automatically post preview comments (with screenshots) as GitHub PR comments
API endpoint: Expose a REST API to list comments for a deployment
The Workflow This Enables
Human reviews preview deployment → spots UI issue → leaves comment with screenshot
Comment syncs to GitHub PR (or is fetched via API)
AI agent in CI picks up the feedback, understands the issue, pushes a fix
New preview deploys automatically
Human verifies → resolves comment
This closes the feedback loop instantly and makes human review time dramatically more valuable - their feedback directly triggers fixes rather than sitting in a queue.
Why Vercel Should Care
This would make preview comments a key integration point in AI-augmented development workflows, increasing their value and stickiness compared to competitors.
This is a really interesting feature request and is a great example of how preview deployments could evolve for AI-driven workflows. I’m flagging this with the team as a concrete use case!
I came here to write a nearly identical request. The markdown export per comment inside the Vercel Toolbar has ll the relevant info for ai coding tools.
The Ask:
Expose a REST API endpoint to retrieve all comments for a deployment, including:
Comment text and metadata (author, created date, status)
Page path and viewport information
DOM selector of the commented element
React component tree context
Deployment URL
The Use Case:
Right now, my team leaves detailed feedback on preview deployments using the toolbar. But when it’s time to implement those changes, we have to manually copy each comment one by one. With programmatic access, we could:
Batch export feedback into our AI-assisted code editor, so developers can see all comments with their exact element targets
Automate documentation of requested changes
Create custom workflows that integrate Vercel feedback directly into our development tools
Reduce context switching by pulling all actionable feedback into a single view
Why This Matters:
Comments are already capturing the hard part (precise element targeting + component context). Making this data programmatically accessible would transform Comments from a discussion tool into a powerful integration point for modern dev workflows. Teams using AI-assisted coding, automated testing, or custom CI/CD pipelines would benefit enormously.
Example Integration:
GET /v1/deployments/{deploymentId}/comments
GET /v1/projects/{projectId}/comments?branch={branchName}
This would unlock so much potential for how teams collaborate on Vercel previews.