SiteRelay

Company: SiteRelay

Self-reported basis: The description is entirely from the project author’s own submission to the OpenAI 2026 hackathon on Devpost. No external verification or historical data is available.

What it appears to be: A local-first browser-to-Codex inspection bridge that captures rendered web interfaces and makes their structure, styles, assets, and motion available to AI agents via a Codex MCP plugin.

What changed: The project was built as part of a hackathon submission; no prior version or commercial history is evidenced.

Most important open question: Does SiteRelay have any evidence of adoption, usage, or traction beyond the author’s own demonstration and development?

The description states that SiteRelay is a local-first browser-to-Codex inspection bridge. It consists of:

• A Chrome extension that captures components, sections, or full pages.
• A local relay service that stores the capture data.
• A Codex MCP plugin that allows querying the captured data.
• A React renderer for reconstructing interfaces.
• A visual difference heatmap to compare rendered vs. source.

The tool is described as enabling developers to point at an authorized interface in the browser and let Codex inspect the browser-rendered truth directly, rather than relying on screenshots or large page dumps.

Evidence: The author's own write-up.

Inference: This is a developer tool for AI-assisted web development that focuses on fidelity and local processing.

The description states:

• SiteRelay was built to address limitations of screenshots, which hide details needed by coding agents.
• It aims to enable AI agents to inspect browser-rendered truth directly.
• The tool is positioned as a local-first solution, avoiding third-party uploads.
• It supports fidelity-first reconstruction and visual regression scoring.

The positioning appears to be that SiteRelay is a developer tool for AI workflows, focused on render fidelity, privacy, and local processing.

Evidence: The author's own write-up.

Inference: This is a niche product aimed at developers working with AI agents in web development contexts.

The description states that SiteRelay is intended for developers who work with AI agents (e.g., Codex) to inspect and reconstruct web interfaces.

It is not clear whether the tool targets a specific segment of developers, such as frontend engineers or full-stack developers.

Evidence: The author's own write-up.

Inference: Likely aimed at developers working in AI-assisted web development environments.

There is no evidence of any business model or pricing structure in the description.

The project is described as a hackathon submission with no indication of monetization, licensing, or commercial use.

Evidence: Not evidenced.

Inference: No business model or pricing data available.

The project is built using:

• TypeScript
• A pnpm monorepo
• A Manifest V3 Chrome extension
• A local relay service
• A Codex MCP server and plugin
• A React renderer
• Shared schemas, tests, and automated setup flows for Windows/macOS

The system is described as:

• Local-first, with no third-party data upload
• Secure, with pairing, origin checks, and authorization controls
• Fidelity-focused, with visual difference heatmaps and animation timeline reconstruction
• Automated, with one-command setup and automated tests

Evidence: The author's own write-up.

Inference: The tool is technically sophisticated for a hackathon project, but lacks evidence of production deployment or scalability.

There is no evidence of any traction, customers, or usage beyond the authors’ own development and demonstration.

The project was submitted to a hackathon and has no reported adoption, revenue, or user base.

Evidence: Not evidenced.

Inference: No signs of product-market fit or commercial traction.

There is no evidence in the description of any competitive landscape or direct competitors.

The tool appears to be positioned within a niche space involving AI-assisted web development and browser rendering fidelity, but no comparison to existing tools is made.

Evidence: Not evidenced.

Inference: No competitive context provided; cannot assess positioning relative to other tools.

• No commercial traction or adoption — the project is a hackathon submission with no evidence of real-world usage.
• Limited team size (2 members) — raises questions about scalability and long-term maintenance.
• Local-first approach may limit utility — if the tool only works locally, it may not integrate well into broader workflows.
• No pricing or monetization model — unclear how this would be commercialized.
• High technical complexity for a hackathon project — while impressive, it’s unclear whether this is sustainable or scalable.

Evidence: Not evidenced.

Inference: Risks are inferred from lack of evidence and the nature of a hackathon submission.

1. What specific use cases or workflows does SiteRelay enable that current tools do not?
2. How is the authorization and provenance model enforced in practice?
3. Are there any plans to support other browsers beyond Chrome?
4. Has the tool been tested with real developers or teams, or is it purely a proof-of-concept?
5. What are the long-term plans for monetization or commercialization?
6. How does SiteRelay handle edge cases like dynamic content or complex animations?
7. Is there any internal testing or feedback from users beyond the authors?