Suede Signal for Alexa+

Inspiration

Business owners need a useful next step from a website assessment. A long technical report can hide the practical question: what should I fix first? Suede Signal packages website-readiness checks into concise responses designed for conversational interfaces.

What it does

Suede Signal for Alexa+ by Jason Colapietro of Suede AI / Suede Labs exposes three read-only Model Context Protocol (MCP) tools: website assessment, two-site comparison, and explanation of an individual scoring area. The tools return scores, weaknesses, and concrete improvements across crawler access, metadata, structured data, citability, and trust. These are technical SEO and AI search visibility readiness checks; the scores do not measure actual AI-engine mentions, rankings, or referrals.

How we built it

The TypeScript server uses the MCP SDK, Zod input schemas, and Streamable HTTP. Its deployed endpoint negotiates MCP 2025-11-25 and calls the existing Suede Signal audit API. Assessment and comparison include schema-backed structured results; the area-explanation tool returns text. A speech-formatting layer keeps recommendations concise and suitable for reading aloud. Vercel hosts the endpoint.

This is a self-hosted MCP implementation for the Alexa+ track. The demo shows real initialization, discovery, and calls to all three tools. A native Alexa+ device or preview SDK integration is not demonstrated. The current server does not edit websites or execute remediation tasks.

Existing work and hackathon contribution

The Suede Signal audit engine and API predate this entry. The hackathon work adds the MCP server, three tools, Streamable HTTP endpoint, speech-formatting layer, competitor comparison, area explanations, and tests. The inspected implementation commits date from October 12–13, 2026.

Verified result and engineering challenges

Live verification negotiated the required protocol, discovered all three tools, and successfully called check_ai_visibility, compare_ai_visibility, and explain_ai_visibility_area. A key design challenge is translating page checks into useful conversational recommendations while distinguishing technical readiness scores from actual search outcomes. The implementation keeps actions read-only and formats practical fixes. The README documents 19 tests; a fresh test-suite run is not claimed here.

Try it

Source and setup: https://github.com/JasonColapietro/suede-signal-alexa-mcp MCP endpoint: https://suede-signal-alexa-mcp.vercel.app/mcp Use an MCP client supporting Streamable HTTP, initialize with protocol 2025-11-25, list tools, and call the website assessment, comparison, and area-explanation tools. No authentication is required for the endpoint.