Audio Intelligence Mcp

Transcribe, summarize, and analyze audio files using local Whisper + Qwen. Returns transcript, segments, and action items.

Security Scan

The skill claims 100% local inference, but its code posts audio to external ntriq servers and performs micropayment/charging behavior that is not reflected in the declared requirements—this mismatch is a significant coherence and privacy concern.

Purpose & Capability

The name/description claim local Whisper + Qwen inference ("100% local AI inference (zero external API calls)"), but the code calls an external server (default SERVER = https://ai.ntriq.co.kr or process.env.NTRIQ_AI_URL) to perform transcribe/summarize/analyze. The README and SKILL.md also reference external endpoints (x402.ntriq.co.kr) and micropayments, while the registry metadata declares no required env vars or credentials — this mismatch indicates the implementation does more (remote processing and payments) than the description implies.

Instruction Scope

SKILL.md and README emphasize local processing and privacy, but the runtime handlers send POST requests with audio URLs to external endpoints (/audio/transcribe, /audio/summarize, /audio/analyze). The SKILL.md example also references an external payment flow. The instructions do not disclose that audio (audioUrl) will be forwarded to a remote service, so users may unknowingly expose audio content and metadata to third-party servers.

Install Mechanism

No install spec is provided (instruction-only at registry level), which is lower install risk. The package includes typical npm dependencies (apify, express, @modelcontextprotocol/sdk). There are no downloads from arbitrary URLs or extract/install steps in the manifest. Running the code would require npm install; nothing in the install metadata indicates hidden installers.

Credentials

The manifest lists no required env vars or credentials, but code uses process.env.NTRIQ_AI_URL (optional) and Apify APIs (Actor.charge, Actor.init) which rely on Apify environment variables (e.g., APIFY_TOKEN) when charging. README also instructs embedding an APIFY token in MCP URLs. The skill requests charging behavior (micropayments) without declaring any credential needs, which is disproportionate and under-documented.

Persistence & Privilege

The skill is not marked always:true and does not attempt to modify other skills or system-wide settings. It logs and attempts to charge via Apify Actor but does not persist configuration into other skill configs. No elevated persistence privileges are requested in the manifest.

What to consider before installing

This skill is inconsistent: it promises local, private inference but the code forwards audio URLs to external ntriq servers and contains micropayment/charge logic. Before installing, verify with the author where audio is actually processed (local vs remote), whether audio/data are retained, and which account/token is used for charges. Do not send sensitive audio to this skill until you confirm the remote endpoints and privacy policy; if you must test, do so with non-sensitive audio in a sandbox and monitor network traffic. Also ask the publisher to declare required environment variables (e.g., APIFY token) and to reconcile the differing domains (x402.ntriq.co.kr vs ai.ntriq.co.kr) and the "local" inference claim.

Audio Intelligence

Transcribe, summarize, and analyze audio files using local Whisper + Qwen. Returns transcript, segments, and action items.

Usage

x402 Payment (AI agents)

curl -X POST https://x402.ntriq.co.kr/audio-intel \
  -H "Content-Type: application/json" \
  -d '{"image_url": "https://example.com/doc.png"}'
# Returns 402 → auto-pay USDC → get result

Service Catalog

curl https://x402.ntriq.co.kr/services

Features

• 100% local AI inference (zero external API calls)
• x402 micropayments (USDC on Base)
• Sub-10 second processing
• JSON structured output

Powered by

• ntriq Data Intelligence
• Qwen2.5-VL (vision) + Qwen2.5 (text) local inference