speak-global-mcp 0.1.2

MCP server for the SpeakGlobal Developer API (/v1).

speak-global-mcp

Connect desktop AI apps and editors that support the Model Context Protocol (MCP) to SpeakGlobal: voice-style TTS, recordings, transcription, and audio processing via the same Developer API (/v1) as your dashboard key (Account → Developer).

Install	PyPI: speak-global-mcp — pip install / pipx install
Source & issues	github.com/Design-Aura/speak-global-MCP

Limits: Export billing APIs are not exposed as MCP tools (/v1/tts/export, export-status, /v1/recordings/export).

---

Requirements

• Python 3.10+
• A SpeakGlobal Developer API key (sg_live_…) from the dashboard

---

Install

From PyPI (recommended)

pipx (keeps dependencies separate — good for CLI-style tools):

pipx install speak-global-mcp
pipx ensurepath

Open a new terminal so speakglobal-mcp is on your PATH.

Pin a version if you want (see PyPI versions):

pipx install "speak-global-mcp==X.Y.Z"

venv + pip:

python -m venv .venv
.venv\Scripts\activate          # Windows
# source .venv/bin/activate     # macOS / Linux
pip install speak-global-mcp

---

Connect an MCP client

This server speaks MCP over stdio: your app starts speakglobal-mcp (or python -m speak_global_mcp.server) locally. Put the API key only in that app’s config — never commit it.

Most clients use the same shape: a top-level mcpServers object with one entry per server: command, optional args, and env for SPEAKGLOBAL_*.

Use SPEAKGLOBAL_ENVIRONMENT=production with the live API (HTTPS + non-empty key). Restart or reload MCP in the app after editing config.

Example: speakglobal-mcp on PATH (after pipx)

{
  "mcpServers": {
    "speakglobal": {
      "command": "speakglobal-mcp",
      "env": {
        "SPEAKGLOBAL_ENVIRONMENT": "production",
        "SPEAKGLOBAL_API_BASE": "https://api.speakglobal.ai",
        "SPEAKGLOBAL_API_KEY": "sg_live_REPLACE_ME"
      }
    }
  }
}

Example: Python from a venv (command not on PATH)

Use your real interpreter path. On Windows, escape backslashes in JSON or use forward slashes.

{
  "mcpServers": {
    "speakglobal": {
      "command": "C:/path/to/.venv/Scripts/python.exe",
      "args": ["-m", "speak_global_mcp.server"],
      "env": {
        "SPEAKGLOBAL_ENVIRONMENT": "production",
        "SPEAKGLOBAL_API_BASE": "https://api.speakglobal.ai",
        "SPEAKGLOBAL_API_KEY": "sg_live_REPLACE_ME"
      }
    }
  }
}

Where to paste it (by app)

App	Config file (typical location)
Cursor	Windows: %USERPROFILE%\.cursor\mcp.json — macOS/Linux: ~/.cursor/mcp.json
Claude Desktop	Windows: %APPDATA%\Claude\claude_desktop_config.json — macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windsurf	MCP settings in the editor (often under AI / MCP); same mcpServers JSON — see Windsurf docs for the exact file or UI path.
VS Code (MCP extension)	Use the extension’s MCP configuration (command + env); format matches the examples above where the extension expects stdio servers.
Other MCP hosts	Any product that supports stdio MCP servers: use the same command / args / env pattern and follow that product’s documentation for where config lives.

Remote MCP over HTTP (sse / streamable-http) is optional for self-hosting; see CONTRIBUTING.md. Many desktop apps only support local stdio.

---

Voice clone & TTS

The server does not create voice profiles. Create or clone a voice in the web app first, for example:

https://speakglobal.ai/dashboard/tts/create-profile

Then use MCP tools such as list_voice_profiles and generate_tts with the profile’s session_id (or your account default when set).

---

What the tools do

Tool	What it’s for
account_summary	Account, credits, default voice session
list_voice_profiles / get_voice_profile	Voice profiles for TTS
list_recordings / get_recording	Library & detail
generate_tts	Text → speech (voice clone)
presign_audio_upload	Upload new audio (then PUT file, register URL)
register_recording_from_url	Register recording from a URL
start_transcription	Transcribe
start_enhance / edit_transcript / start_re_enhance	Processing pipeline
get_job_status / list_jobs_for_session	Job tracking

API scopes apply to your key — use a key with only the scopes you need.

---

Optional: Docker (self-hosted MCP over HTTP)

For SSE / streamable-http behind your own TLS, you can run the container (see CONTRIBUTING.md for defaults and env vars). Do not expose it publicly without TLS and access control.

---

Security

• Treat SPEAKGLOBAL_API_KEY like a password; rotate it if leaked.
• One running server typically uses one key (one SpeakGlobal account).

---

Contributing & development

Clone, tests, release process, full env reference, Docker, and CI: CONTRIBUTING.md.