Echoteka for your AI assistant

What it is

MCP is an open protocol that lets an AI assistant call tools. Echoteka's server exposes the mastering lab as nine tools. You give the assistant a public link to a track; it measures it, suggests a preset and a streaming target, and starts a master. You listen and download the result. Nothing is installed on your computer: the server is remote and the audio is fetched by Echoteka from the link you provide.

Server address (streamable HTTP, no key needed): https://echoteka.ai/mcp

Connect it

Claude Code

claude mcp add --transport http echoteka https://echoteka.ai/mcp

Claude Desktop

Open Settings, then Connectors, add a custom connector and paste https://echoteka.ai/mcp. If your version only supports local servers, use the bridge: add this to claude_desktop_config.json and restart the app.

{
  "mcpServers": {
    "echoteka": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://echoteka.ai/mcp"]
    }
  }
}

Cursor

Add this to ~/.cursor/mcp.json (or .cursor/mcp.json in a project):

{
  "mcpServers": {
    "echoteka": { "url": "https://echoteka.ai/mcp" }
  }
}

OpenAI Codex

In a terminal:

codex mcp add echoteka --url https://echoteka.ai/mcp

Or add this to ~/.codex/config.toml:

[mcp_servers.echoteka]
url = "https://echoteka.ai/mcp"

The Codex CLI, the Codex IDE extension and the ChatGPT desktop app share this configuration, so one setup covers all three.

Any other client

Send JSON-RPC over HTTP POST to https://echoteka.ai/mcp. Start with server/discover or tools/list. A machine-readable description is available at /mcp.json.

Sign in from your assistant

Anonymous use is enough to try the tools. To use the masters of your own account, add the server with the sign-in address instead:

claude mcp add --transport http echoteka https://echoteka.ai/mcp/oauth

In Claude Desktop or any client with a connector dialog, paste https://echoteka.ai/mcp/oauth. The client opens an Echoteka page in your browser: sign in or create an account, press Allow, and you are connected. The sign-in uses OAuth 2.1 with PKCE; the assistant gets an access token limited to mastering and never sees your password. Changing your password disconnects every app.

What you get after sign-up: 3 welcome masters, available once you have confirmed your email, plus 1 free master every day. They are used through the assistant exactly as in the web lab. The assistant still cannot buy credits: if the masters run out, it tells you and you decide on the website.

The nine tools

What is free

• Analysis and artifact cleanup are free and do not use your daily master.
• One full master every day with no account: one per device or network, days counted in UTC. It is a complete master of the whole track.
• 3 welcome masters after you sign up and confirm your email, usable through the assistant when you sign in, plus 1 free master a day.
• Previews of 24 seconds in several presets: 6 per day.
• Extra masters use credits. An assistant cannot pay, and the server has no tool for buying credits.

The same quotas and checks apply as on the website. Using several devices, networks or accounts to collect more free masters is against the terms. Free masters are not stored for long, so download the result when the job is done.

Honest limits

• Links, not uploads. The assistant passes a public https link to the audio file; Echoteka fetches it. Share links for Google Drive, Dropbox and similar pages are resolved to the file when possible. Up to 100 MB per file, 10 tracks and 300 MB per batch.
• Sign-in is optional. Without it the server uses the guest free master. Connect through https://echoteka.ai/mcp/oauth and sign in to use your account's masters instead; see sign in from your assistant. The assistant never receives your password.
• Cleanup is not a disguise. It removes artifacts a generator leaves in the audio. It does not change where a track came from, and detectors that judge timbre still recognise AI music. If a platform or distributor asks you to disclose AI use, disclose it. See AI artifact cleanup.
• Mastering is not mixing. It cannot fix a vocal that sits too low or a muddy mix.