MCP server
The Munsit MCP server lets AI apps such as Claude and ChatGPT use your Munsit account: find a voice, generate speech, transcribe a recording or check your credits, just by asking. You sign in with your Munsit account — there is no API key to copy.
Connect it
Use the URL for the region your account is in — the same region as the Munsit app you sign in to.
| Your Munsit app | MCP server URL |
|---|---|
| app.munsit.com | https://mcp.munsit.com/mcp |
| sa.app.munsit.com | https://sa.mcp.munsit.com/mcp |
| ae.app.munsit.com | https://ae.mcp.munsit.com/mcp |
Sign-in and permissions
When you connect, the app opens a Munsit sign-in page and asks you to approve access. Only approve apps you trust. Each permission unlocks a group of tools; you stay connected until you remove the app.
| Permission | What it allows |
|---|---|
account:read | Your plan, credit balance, usage, credit history and the names of your API keys. Key secrets are never shown. |
history:read | Your speech and transcription history, and your transcripts. |
voices:read | The voice library, your cloned voices, and the models with their prices. |
tts | Generate speech (spends credits) and add tashkīl. |
stt | Upload and transcribe audio (spends credits). |
offline_access | Stay connected: the app renews its access without asking you to sign in again every day. |
If a permission is missing, the tools that need it ask you to reconnect and approve it.
What the app can do
Thirteen tools, covering what the Munsit app does. Only two spend credits.
| Tool | What it does | Credits |
|---|---|---|
| get_account | Plan, credit balance, auto top-up and limits. | Free |
| get_usage | Credits spent and volume over a date range, per product. | Free |
| list_credit_transactions | Purchases, top-ups, refunds and usage charges. | Free |
| list_api_keys | Names and masked endings of your API keys. | Free |
| list_activity | Your recent speech and transcriptions, with fresh audio links. | Free |
| get_transcript | A transcript as text, SRT or VTT. | Free |
| search_voices | Find voices by dialect, gender, age, language or description, in English or Arabic. | Free |
| get_voice | One voice's full details. | Free |
| list_models | Speech and transcription models with their prices. | Free |
| text_to_speech | Generate speech from Arabic or English text and return a download link. | Characters × the model's rate |
| diacritize_text | Add tashkīl for Fusha voices. | Free — the marks count as characters in text_to_speech |
| create_upload_url | Upload a local audio file, for apps that can run commands, such as Claude Code. | Free |
| transcribe_audio | Transcribe a recording from a link or an upload. | Per minute of audio |
Prices are the same as in the app — ask the app to list the models to see them. 20,000 credits = 1 USD.
Try it
Ask in English or Arabic. The app picks the tools.
Credits and limits
| Limit | Value |
|---|---|
| Speech per request | Up to 10,000 characters |
| Transcription per file | Up to 500 MB and 60 minutes. Long recordings keep transcribing in the background — ask for the transcript when it is done. |
| Audio links | Valid for 1 hour. Ask for your recent activity to get a fresh link. |
| Repeated speech | The same speech request within 15 minutes returns the same audio without charging again. |
| Requests per minute | 120 for reading, 20 for speech and transcription, 30 for tashkīl, 10 for usage reports. |
| Auto top-up | If you have auto top-up on, activity from Claude can trigger it at most once a day. Activity from other apps never triggers it: they can only spend your current balance. |
Your plan and region access apply as in the app. If you run out of credits, the app tells you and links to billing; nothing is charged for the request it could not run.
Disconnect and privacy
Remove Munsit in the app's settings (in Claude: Settings → Connectors), or run claude mcp remove munsit in Claude Code. Access already granted can last up to a day after you disconnect.
The app only receives what a tool returns in your conversation — for example a transcript or a voice's details. It never receives your API key secrets or payment details.
Working with an AI assistant? Every page is available as Markdown: add .md to the URL, or send an Accept: text/markdown header. For the whole documentation in one request, point it at llms-full.txt; the page index is llms.txt. Or use Copy Page, top right.