API Documentation

RESTful API with simple authentication

MCPttsaudit-mcp

The TTSAudit MCP server lets AI agents - Claude, Cursor, and any Model Context Protocol client - run audio audits directly. Point an agent at a folder of Text-to-Speech files and it can score them, flag which tracks to regenerate, and check your credit balance, all without you touching the dashboard.

It runs locally over stdio and reads audio straight from your machine, then calls the same /audit API you would use directly. That is why it can upload files an agent could not otherwise send: the server reads them off disk for you.

Step 1: get your API key

You'll paste it into the config below. Create one under Dashboard → Settings.

Get API key

Install

Requires uv (or Python 3.10+). The package runs straight from PyPI - no clone needed.

terminal
uvx ttsaudit-mcp

You will need an API key. Create one in your dashboard settings. Each enabled check costs 1 credit per file (cached files are discounted). See pricing.

Connect your client

claude mcp add ttsaudit \
  --env TTSAUDIT_API_KEY=sk_your_key_here \
  -- uvx ttsaudit-mcp

Replace sk_your_key_here with your API key, then restart the client. Once connected, ask the agent something like "audit the wav files in ./output for quality and pacing."

Tools

The server exposes four tools. The agent decides when to call them.

ToolDescription
audit_ttsRun quality, speaker consistency, pacing, and script-accuracy checks on local audio files. Returns per-file scores and the list of tracks to regenerate.
get_balanceReturn the remaining credit balance for the configured API key.
list_auditsList recent audits for the account, newest first, with pagination.
get_auditFetch the full results of a past audit by its auditId.

Configuration

Set via environment variables in your client config.

VariableRequiredDescription
TTSAUDIT_API_KEY Required Your API key from the dashboard. Used to authenticate every request.
TTSAUDIT_BASE_URL Optional Override the API base URL. Defaults to https://api.ttsaudit.com/v1.

Example session

A typical exchange once the server is connected.

You

Audit the three narration files in ./chapters for quality and speaker consistency.

Agent

Calls audit_tts with the three files and the quality + comparison checks, then reports:

Overall score 82/100. chapter3.wav is flagged for regeneration - its voice similarity deviates 18% from the others. The other two passed. 6 credits used; 38 remaining.

Prefer calling the API directly? See the REST API reference. The MCP server is a thin wrapper over the same endpoints.

Ready to integrate?