PRGate

PRGate is a self-reported tool that audits live websites for accessibility issues, proposes code fixes in isolated branches, re-renders and verifies those changes via browser automation, and opens GitHub pull requests only when fixes are proven to work. The system uses AI (GPT-5.6 Terra) for patch planning and verification, with Playwright and axe-core for browser-level evidence capture. It is built as a SaaS-like dashboard using Next.js, React, Supabase, and Inngest, with backend services running on Render and Vercel.

The description states that PRGate aims to close the gap between accessibility tooling identifying issues and teams being able to prove fixes work in real user contexts. It is positioned as a solution for developers or accessibility consultants who want to automate remediation while maintaining human review and proof at the center.

Key open question: Does PRGate actually function as described, or is this a conceptual prototype? There is no evidence of actual deployment, usage, or customer feedback beyond the hackathon submission.

The description states that PRGate audits live public previews for accessibility barriers using axe-core and visual analysis. It captures browser screenshots and accessibility-tree evidence, proposes source-level fixes in isolated GitHub branches, runs repository tests, waits for Vercel Preview Deployments, re-audits the changes, stores before/after evidence, and opens a GitHub pull request only when all fixes verify.

PRGate is described as using Playwright and axe-core for browser automation and static analysis. It uses GPT-5.6 Terra (via Codex) for patch planning and verification. The system runs on Docker via Render, with Vercel Sandbox for isolated code changes and deployments. Supabase handles authentication, run history, and rate limiting. Inngest orchestrates workflows including audit, patch, preview, verification, daily rescan, and retention.

The product is described as a dashboard built with Next.js, React, and TypeScript, hosted on Vercel, with GitHub OAuth integration.

The description states that PRGate was inspired by the question: "what if an accessibility agent could inspect a rendered website, fix the underlying code safely, look at the deployed result again, and only claim success when it has evidence?"

It positions itself as not replacing accessibility experts but removing repetitive remediation work while keeping human review and proof central. The authors claim that PRGate completes a real closed loop instead of stopping at reports or diffs.

The project evolved from a hackathon submission to a concept that aims to make autonomous coding valuable only when paired with strict verification, rather than simply producing more code. It is described as aiming for a continuous accessibility reliability layer that proves fixes remain valid over time.

The description states that PRGate is intended for developers or accessibility consultants who want to automate remediation while maintaining human review and proof at the center. It targets teams working with websites that need accessibility improvements, particularly those struggling with backlog management and proving fixes work in real user contexts.

It appears to be aimed at organizations using GitHub repositories, Vercel deployments, and public preview environments. The system is designed for use with web applications built on modern stacks (React, Next.js) and deployed via Vercel.

Not evidenced.

The description does not contain any information about pricing models, monetization strategies, or business model details beyond the self-reported project scope.

PRGate is described as using Playwright and axe-core for browser automation and static analysis. It uses GPT-5.6 Terra (via Codex) for patch planning and verification. The system runs on Docker via Render, with Vercel Sandbox for isolated code changes and deployments. Supabase handles authentication, run history, and rate limiting. Inngest orchestrates workflows including audit, patch, preview, verification, daily rescan, and retention.

The architecture separates the lightweight dashboard from the browser-capable worker to improve reliability and security boundaries. It uses Vercel Sandbox for isolated environments, runs repository tests before deployment, and employs bounded retries with human review fallbacks.

Not evidenced.

There is no evidence of revenue, customers, usage metrics, or adoption beyond the hackathon submission. The project is described as a prototype built for a hackathon event.

Not evidenced.

The description does not mention any competitors or competitive landscape. No information is provided about existing tools in the accessibility automation space or how PRGate differentiates from them.

• The system appears to be a hackathon prototype with no evidence of real deployment or usage.
• The use of GPT-5.6 Terra (noted as "GPT-5.6 Terra" in the tech stack) is unusual and may indicate an unverified or fictional model name.
• The architecture involves complex integrations between multiple services (Render, Vercel, GitHub, Supabase, Inngest) that may not be fully functional or scalable.
• There is no evidence of actual customer feedback, usage data, or performance metrics.
• The system's reliance on browser automation in serverless environments raises questions about reliability and scalability.
• The claim of "real closed loop" execution without any verification of actual functionality.

1. What is the actual functional status of PRGate? Is it currently operational or still a prototype?
2. How does PRGate handle edge cases in browser automation that might not be covered by the current implementation?
3. What are the actual constraints and limitations of the system's current architecture?
4. Can you demonstrate end-to-end functionality of the closed-loop process described?
5. What is the current state of integration with different CI/CD pipelines and deployment environments?
6. How does PRGate handle large or complex repositories that might not fit within sandboxed environments?
7. What are the actual costs associated with running PRGate at scale?
8. How do you plan to address provider quotas, model failures, and other operational challenges in a production environment?