Brainiall MCP Server

What fasuizu-br/brainiall-mcp-server MCP server does

The fasuizu-br/brainiall-mcp-server MCP server connects MCP clients to the Brainiall Speech AI API at https://api.brainiall.com. It provides four speech-focused tools: pronunciation scoring, speech transcription, speech synthesis, and voice discovery.

Pronunciation assessment accepts reference text and base64-encoded WAV or MP3 audio. The result includes an overall score from 0 to 100, along with feedback at word and phoneme level. Transcription accepts base64-encoded WAV, MP3, WEBM, or OGG audio and can detect the spoken language automatically. Text synthesis returns MP3 speech, while voice listing returns available voice IDs and supported languages.

How it works

The MCP server receives tool calls from a compatible client and forwards the relevant request to Brainiall’s Speech AI API. Audio is supplied inline as a base64 string rather than as a file path or upload reference. The API key is read from the BRAINIALL_API_KEY environment variable.

assess_pronunciation requires the text the speaker was expected to read and the recorded audio. Its optional language value defaults to en-US. transcribe_speech accepts an optional language hint, while synthesize_speech supports a voice ID and a speed multiplier from 0.5 to 2.0. Its default voice is alloy, and its default speed is 1.0. list_voices does not require parameters.

Setup and configuration

To use the fasuizu-br/brainiall-mcp-server MCP server, obtain an API key from app.brainiall.com, then set BRAINIALL_API_KEY in the server process environment. The default local mode uses stdio, which is suitable for clients such as Claude Desktop and Cursor.

The server can also run with Streamable HTTP on port 8000. In that mode, an MCP client connects to http://localhost:8000/mcp. Docker usage is documented as an alternative: build the image locally, pass the API key into the container, and publish port 8000.

Tools and capabilities

• assess_pronunciation: Scores spoken audio against supplied reference text and reports word- and phoneme-level feedback.
• transcribe_speech: Converts supported audio formats to text and performs automatic language detection.
• synthesize_speech: Converts text to MP3 audio using a selected voice and playback speed.
• list_voices: Returns the available text-to-speech voices and their supported languages.

Limitations and notes

The server depends on access to the external Brainiall Speech AI API and cannot operate without a valid BRAINIALL_API_KEY. The documented audio inputs must be encoded as base64 before being passed to the tools. Pronunciation assessment accepts WAV and MP3, while transcription additionally supports WEBM and OGG. The README identifies the project as MIT-licensed and does not document usage quotas, latency, billing beyond the availability of a free API key, or offline processing.