What can AI agents do with ElevenLabs through Ferentin?

Through Ferentin, agents can use ElevenLabs for voice agent management, conversation review, text to speech and usage cost estimation. Every call is checked against your Ferentin policies first.

How is ElevenLabs access secured?

Requests through the ElevenLabs integration are covered by OAuth 2.0 user consent, cimd client identity, PKCE s256, scoped access and data residency regions. These controls are applied by Ferentin, not by ElevenLabs, so they hold across every connected agent.

What permissions does the ElevenLabs integration require?

The ElevenLabs integration requests 7 OAuth 2.0 scopes: convai_read (read access); convai_write (elevated access); text_to_speech; speech_history_read (read access); flows; image_video_generation; voice_generation

How do I connect ElevenLabs to Ferentin?

Open the Ferentin admin console, add ElevenLabs as a new connection and authenticate with OAuth 2.0. You then choose which permissions and which tools your agents may use, and the connection goes live for everyone covered by your policies.

Is the ElevenLabs integration generally available?

The ElevenLabs integration is currently beta on Ferentin.

ElevenLabs

Beta

Build and operate ElevenLabs voice agents — manage agents, review conversations, and generate speech.

ElevenLabs
Voice
Text to Speech

Configure ElevenLabs

Authentication: OAuth 2.0
Hosting: Vendor-hosted endpoint
Required scopes: 7
Data actions: Read and write

How it works

Every agent request passes through Ferentin policy before it reaches ElevenLabs.

Ferentin connects each user to their own ElevenLabs workspace over OAuth 2.0. ElevenLabs supports hosted client metadata (CIMD) and not dynamic client registration, so Ferentin presents its published client-metadata URL as the client_id — no client secret is stored. There is no agent mode: the server accepts OAuth user tokens only, so every tool call is attributable to a consenting user, and the scopes granted at connect time bound what it can do.

AI agentAsks for something
Ferentin policies
- OAuth 2.0 user consent
- Cimd client identity
- PKCE s256
- Scoped access
ElevenLabs MCP serverAuthorized connection
Governed resultMasked, logged, returned

What you can do with ElevenLabs

Voice agent management
Conversation review
Text to speech
Usage cost estimation

Permissions requested

The OAuth 2.0 scopes ElevenLabs asks you to approve when you connect.

Scope	Access
convai_read	Read
convai_write	Elevated
text_to_speech	Not rated
speech_history_read	Read
flows	Not rated
image_video_generation	Not rated
voice_generation	Not rated

Before you connect

OAuth only. The hosted MCP rejects an xi-api-key header and rejects an ElevenLabs API key sent as a bearer token, both with 'OAuth bearer token required for the hosted MCP' — so this server has no agent mode and cannot run unattended on a tenant credential.
ElevenLabs does NOT support Dynamic Client Registration on this server: its authorization server metadata advertises client_id_metadata_document_supported and publishes no registration_endpoint. Ferentin therefore uses CIMD, which requires mcp.cimd.client-id-url to be configured on admin-api-server.
The OAuth endpoints in this card are the global (US) ones. Every residency region is a SEPARATE authorization server with its own authorize, token and revoke endpoints, and Ferentin does not re-discover them from an overridden server URL — a regional install that changes only the server URL will still be sent to the US authorization server.
The protected-resource metadata for the global endpoint names https://api.us.elevenlabs.io/v1/mcp as the resource, not the api.elevenlabs.io host you dial. This card does not send an RFC 8707 `resource` parameter; enable one only after confirming the audience ElevenLabs actually mints.

2 more technical notes

Scopes cover far more than agent management — voice generation is voice cloning, and image/video generation produces synthetic media. Both are off by default here; grant them deliberately.
Tools are not enumerated in this card. The server exposes no tool list to an unauthenticated caller, so the surface is populated on first authenticated discovery rather than asserted here.

About ElevenLabs

ElevenLabs is an AI audio platform for speech synthesis, voice cloning, and ElevenAgents — conversational voice agents that handle phone and in-product conversations. Its hosted MCP server exposes an ElevenLabs workspace to agents over Streamable HTTP.

Website Developer docs

Frequently asked about ElevenLabs

What can AI agents do with ElevenLabs through Ferentin?: Through Ferentin, agents can use ElevenLabs for voice agent management, conversation review, text to speech and usage cost estimation. Every call is checked against your Ferentin policies first.
How is ElevenLabs access secured?: Requests through the ElevenLabs integration are covered by OAuth 2.0 user consent, cimd client identity, PKCE s256, scoped access and data residency regions. These controls are applied by Ferentin, not by ElevenLabs, so they hold across every connected agent.
What permissions does the ElevenLabs integration require?: The ElevenLabs integration requests 7 OAuth 2.0 scopes: convai_read (read access); convai_write (elevated access); text_to_speech; speech_history_read (read access); flows; image_video_generation; voice_generation
How do I connect ElevenLabs to Ferentin?: Open the Ferentin admin console, add ElevenLabs as a new connection and authenticate with OAuth 2.0. You then choose which permissions and which tools your agents may use, and the connection goes live for everyone covered by your policies.
Is the ElevenLabs integration generally available?: The ElevenLabs integration is currently beta on Ferentin.