INTEGRATIONS / MCP
MCP for voice cloning and text to speech
Bring your voices into your AI workflow. The Vocloner MCP server lets a compatible client browse your library, generate voiceovers and clone voices you have permission to use.
- Explore
list_voicesget_usage - Direct your script
- Generate
generate_speech - MP3 / WAV
- Download
get_audio
Your account. Your voices. Your permissions.
Access depends on service activation. If MCP Connections is missing from your account, contact support before configuring a client.
Connect your client. Keep control.
MCP (Model Context Protocol) connects an assistant to tools from other services. In Vocloner, these tools use your account and quotas; the integration does not require keys from other AI providers.
Open Account → MCP Connections and use the service URL shown there. Your client must support Streamable HTTP and OAuth or a Bearer token. Compatibility depends on the client and version.
- OAuth: review the destination and permissions in your browser before authorizing. A client’s declared name does not verify its identity.
- API key: create a key with the permissions you need and save it when displayed. Keys created in Account last 30 days and can be revoked.
- Store the secret in your client environment or secret manager. Do not paste it into chats or public files.
From your library to a voiceover
Available tools depend on the permissions you grant. Start with read access and enable generation or cloning when you need them.
- list_voices · Browse your voices and available public voices.
- get_usage · Check your plan, characters and available voice slots.
- clone_voice · Create a private voice with explicit human confirmation of permission.
- generate_speech · Turn a script into MP3 or WAV using an allowed voice.
- get_audio · Get a download URL that requires your credential.
- get_operation · Check the result of an existing operation.
A practical workflow for your scripts
Ask your assistant to check available voices and usage before generating. Choose a voice, review the script and authorize the voiceover. Then download the result using a client that can send your credential in the Authorization header.
Each generation or cloning request uses a unique operationId. After a timeout, query get_operation with the same identifier before retrying: creating a new identifier can repeat the work and consume quota.
1. list_voices + get_usage
2. generate_speech (voiceId, text, format, operationId)
3. get_operation (operationId)
4. get_audio (generationId)Clear limits. The same usage rules.
Generation keeps your plan limits, moderation and voice access checks. The integration does not expose administration, billing, deletion or voice publication.
MCP cloning accepts recordings between 5 and 270 seconds, up to 2 MiB decoded. It receives base64 audio, not remote URLs or file paths. Use the web upload for larger samples.
- Cloning consent must come from an authorized human; an attached file is not permission.
- Audio downloads require authentication and ownership checks.
- Expired or revoked credentials no longer provide access to tools or audio.
FAQ
Before you start
Does MCP have a separate voice quota?
No. Tools use your Vocloner account and plan limits. Check get_usage or the pricing page before generating.
Does it work with every AI assistant?
You need a client with Streamable HTTP and compatible authentication. Clients supporting only stdio or legacy SSE are not sufficient. Test the connection in your client version.
Can I publish a voice through MCP?
No. clone_voice creates private voices. Publication is handled on the website with its authorization and review requirements.