Tera Studio’s MCP server is at https://terastudio.co/mcp. It gives an AI assistant three read-only tools: find a song’s key and tempo, move a song to a key that fits your voice, and get the steps and link to hear a song in your own voice on terastudio.co. There is no sign-in and no API key. Songs are worked on in your browser and never uploaded.
Connect it
| Server URL | https://terastudio.co/mcp |
|---|---|
| Transport | Streamable HTTP (POST, JSON responses) |
| Authentication | None. No account, sign-in, OAuth or API key. |
- Claude: add a custom connector and paste the server URL. Leave any OAuth fields empty: there is no sign-in step.
- ChatGPT: in developer mode, create an app with the same URL and no authentication.
- Any other MCP client that speaks Streamable HTTP: point it at the server URL.
The three tools
All three are marked read-only, non-destructive and idempotent, and they don’t reach other sites. None of them creates an account, a song or anything else.
Find a song’s key and tempo
find_song_key_and_tempo. Opens a small Tera Studio tool where you choose a song file. It shows the key (major or minor, with its relative key) and the tempo in BPM, with the half-time and double-time readings.
Inputs: None.
Change a song’s key
shift_song_key. Opens the same tool on its key changer. You choose a song file and hear it moved up or down by up to 6 semitones at the same speed, so it fits your voice.
Inputs: semitones (optional): a whole number from −6 to 6; negative is lower. Left out, you choose in the tool.
Start a song in my own voice
start_song_in_my_voice. Returns three steps for hearing a song sung in your own voice on terastudio.co and a link to start there. It creates nothing.
Inputs: song_title, language and occasion (all optional, short text).
The first two open an interactive tool in the chat (an MCP App) where you choose a song file: MP3, M4A or WAV, up to 8 minutes. The file is read and changed in your browser, on your device; it is never sent to Tera Studio or to the assistant. The tool tells the assistant only the reading, such as “C major, about 131 BPM” or how far you moved it, so it can talk about it with you; never the file or its name. In an assistant that doesn’t show interactive tools, the tool answers in text with a link to the same tool on our site: the BPM and key finder or the key changer.
Example prompts
| You ask | Tool the assistant uses |
|---|---|
| “What key and BPM is this song? I have the MP3.” | find_song_key_and_tempo |
| “This song is too high for me. Can you bring it down 2 semitones?” | shift_song_key with semitones −2 |
| “I want to sing this song but it’s too low for my voice.” | shift_song_key (you choose how far in the tool) |
| “I’d love to hear “Happy Birthday” in my own voice for my mum’s birthday.” | start_song_in_my_voice |
What it doesn’t do
- It doesn’t search for, identify or play songs by name. You choose a file you have.
- It doesn’t make songs inside the chat. Songs in your own voice are made on terastudio.co, in your account.
- It only works with your own voice. Tera Studio doesn’t make songs in a celebrity’s or anyone else’s voice.
Privacy
The assistant sends our server the tool it picked and what it filled in, such as how many semitones to move a song or a song’s title. We use that only to answer and don’t keep it or any identifier sent with it: we only count which tool was used and which assistant it came from. Your song stays on your device, and we don’t receive your conversation; the key and tempo the tool passes to the assistant go to the assistant, not to us. The details are in the section “Tera Studio in ChatGPT, Claude and other AI assistants” of our privacy policy.
Technical details
- JSON-RPC 2.0 over Streamable HTTP. Each POST gets one JSON response; there are no sessions and no server-sent event stream.
- Protocol versions 2026-07-28, 2025-11-25, 2025-06-18 and 2025-03-26.
- The interactive tool is the resource
ui://tera-studio/song-key-v1.html(text/html;profile=mcp-app). It makes no network requests. - Up to 120 requests a minute per caller and 64 KB per request. Calls from web pages on other sites are refused.
Support
Email teralabsx@gmail.com or use our contact page. We aim to reply within 24–48 hours.