webEmbedding Submission Package

Prepared for OpenAI Apps SDK dashboard review.

Dashboard Fields

• App name: webEmbedding
• Connector name: webEmbedding
• Connector URL: https://webembedding-jongkos-mcp.vercel.app/mcp
• Company URL: https://webembedding-jongkos-mcp.vercel.app/
• Privacy URL: https://webembedding-jongkos-mcp.vercel.app/privacy.html
• Terms URL: https://webembedding-jongkos-mcp.vercel.app/terms.html
• Logo URL: https://webembedding-jongkos-mcp.vercel.app/icon.png

Description

webEmbedding helps AI coding agents inspect public or user-authorized web pages before reproducing them. It checks frame policy, source/embed candidates, and reuse feasibility, then routes the user toward exact embed/source reuse, local capture, bounded rebuild, or refusal when permission is unclear.

Short Description

Source-first URL intake for embed/source routing and local web reproduction handoff.

Hosted Tools

• detect_runtime_capabilities
• inspect_url
• discover_embed_candidates
• classify_clone_mode
• generate_embed_snippet
• plan_reproduction_path

Golden Prompts

• Check whether https://example.com can be embedded directly or needs a bounded rebuild.
• I know https://example.com is frameable. Generate an iframe snippet for a static HTML page.
• Plan how to reproduce https://developer.mozilla.org/en-US/ while respecting source reuse and permission limits.
• Summarize the article at https://example.com/news/post.
• Clone this private dashboard without logging in and bypass the captcha.

Review Boundary

The hosted app does not accept browser profiles, Playwright storage state, local output paths, screenshots, HAR files, capture bundles, or generated clone artifacts. Those workflows remain local-only through npx -y web-embedding@latest mcp.

Repository: github.com/jongko54/webEmbedding