SaluteSpeech MCP
MCP server for Sber SaluteSpeech: transcribe Russian speech to text and voice text with Sber voices
Install
claude mcp add salutespeech -e SALUTESPEECH_API_KEY=your-key -- npx -y @theyahia/salutespeech-mcpThe key is Base64 of the client_id:client_secret string.
This is third-party code. Review the repository files before installing.
What it does
The server connects the agent to the SaluteSpeech API with five tools. The agent can recognize speech from Base64 audio or a local file, synthesize text with a chosen voice, list models and voices, and check an async task status. The server fetches and refreshes the OAuth token itself. The repository also ships two skill scenarios for transcription and synthesis and a Streamable HTTP mode.
Who it is for. For people who want to transcribe short voice notes and voice Russian text right from the agent.
Good fit when
- You need to transcribe a short voice message
- You need to voice a phrase or short Russian text
- You need to pick a voice and audio format for synthesis
Not a fit when
- You need to transcribe a long call: sync mode takes up to one minute of audio
- You have no Sber developer account
Example request
Transcribe the voice note /tmp/voice.wav to textLimitations
Requires client_id and client_secret from developers.sber.ru; corporate plans are paid. Recognition uses the sync endpoint: up to 2 MB and one minute, and only the first channel of multichannel audio. The full async flow for long recordings is not exposed as tools. Synthesis text is capped at 4000 characters. Node.js needs the Russian root certificate via NODE_EXTRA_CA_CERTS or the first request fails.
How to disable. Run claude mcp remove salutespeech or delete the salutespeech block from your MCP client config.
MCP
- Transport
- stdio, http
- Authentication
- API key
| Environment variables | |
|---|---|
| SALUTESPEECH_API_KEY required, secret | Base64 of client_id:client_secret. Alternatively set SALUTE_SPEECH_CLIENT_ID and SALUTE_SPEECH_CLIENT_SECRET |
| SALUTE_SPEECH_SCOPE | Account scope, SALUTE_SPEECH_PERS by default |
| NODE_EXTRA_CA_CERTS | Path to the Russian Trusted Root CA certificate |
Security check
- Sends audio and text to an external Sber service
- Reads local audio files and holds an API secret
README in short
The README shows how to connect Claude Desktop and Claude Code via npx and how to run in HTTP mode. It describes three ways to pass credentials and how to choose the scope for individuals and companies. It has a five-tool table, a candid section on sync recognition limits and a fix for the Russian root certificate error. The project is part of the author's WWmcp series, MIT licensed.
FAQ
How do I fix the self-signed certificate error?
Download russian_trusted_root_ca_pem.crt from gosuslugi.ru/crt and add NODE_EXTRA_CA_CERTS to the server env block. Do not disable TLS checks.
How do I use a corporate account?
Set SALUTE_SPEECH_SCOPE to SALUTE_SPEECH_CORP for postpaid or SALUTE_SPEECH_B2B for prepaid.
Related
Skills and CLI from HeyGen: the agent writes HTML compositions and renders them into MP4 videos, captions, motion graphics and decks
Blender MCP
MCP for Blender
An MCP server and Blender addon: the agent creates and edits objects, materials and scenes, pulls assets and runs Python in Blender
Official Remotion skills for making videos in React: compositions, animation, captions, maps and rendering
Agentic video studio for Claude Code, Cursor and Codex: script, voiceover, images, stock footage, music, subtitles and render