#Is it possible to add additional languages when creating a new conversational agent via elevenlabs

1 messages · Page 1 of 1 (latest)

blazing socketBOT
#
AI-generated Answer for miky2fois

Yes. When you create an agent via the Agents Create endpoint (used by the JS SDK), you can add additional languages by supplying language_presets in conversation_config and, if you want automatic switching, adding the built‑in language_detection system tool. The docs include an example of creating an agent with language_presets and the language_detection tool, and explain how additional languages behave.

  • Create agent (agents.create) expects conversation_config; you can include language_presets there. (elevenlabs.io)
  • Language detection: shows adding the language_detection system tool and defining language_presets in the agent configuration at creation time. (elevenlabs.io)
  • Language guide: describes adding additional languages and how multilingual mode is used when languages are added. (elevenlabs.io)

So, in short: yes—pass language_presets in conversation_config (and optionally add the language_detection tool) when calling agents.create.

Sources:
Create agent | ElevenLabs Documentation
Language detection | ElevenLabs Documentation
Language | ElevenLabs Documentation

thin topaz
#

And is it possible to add:

  • post call webhook
  • enable overrides
    -enable language detection
blazing socketBOT
#
AI-generated Answer for miky2fois

Short answer:

  • Post-call webhook: Not via agents.create. Post-call webhooks are configured in Agents Platform settings at the workspace level; audio webhooks can be toggled at the agent level via agent webhook overrides in settings. There is no field to set a webhook URL in agents.create. (elevenlabs.io)
  • Enable overrides: You can’t enable the override permissions via API; enable them in the agent’s Security tab (System prompt, First message, Language, Voice ID). Once enabled, you pass overrides at conversation start (conversation_config_override). (elevenlabs.io)
  • Enable language detection: Yes. Add the built-in “language detection” system tool when creating the agent (via tools on the Prompt/Agent config) and make sure the agent has additional languages configured. (elevenlabs.io)

Sources:
Post-call webhooks | ElevenLabs Documentation
Overrides | ElevenLabs Documentation
Language detection | ElevenLabs Documentation

thin topaz
#

How to provide oauth2 to tools when creating an agent

blazing socketBOT
#
AI-generated Answer for miky2fois

Short answer: you don’t pass OAuth2 credentials in the agents.create call. You attach OAuth2 to the tool itself, then reference that tool by ID in the agent.

Do it in three steps:

  1. Create a Workspace Auth Connection for OAuth2
  • In the ElevenLabs dashboard go to Agent settings → Workspace Auth Connections and add an OAuth2 connection (Client Credentials or JWT). This handles token acquisition/refresh for your tools. (elevenlabs.io)
  1. Create the tool and bind it to that Auth Connection

  • Call POST /v1/convai/tools and set api_schema.auth_connection.auth_connection_id to the ID of the Workspace Auth Connection you created. Example request shape is documented on the Create tool endpoint; the response shows the auth_connection structure. Save the returned tool id. (elevenlabs.io)

  • Supported auth methods for tools include OAuth2 Client Credentials and OAuth2 JWT; both are configured via Workspace Auth Connections and then linked to the tool. (elevenlabs.io)

  1. Create the agent and attach the tool by ID
  • In your agents.create payload, include the tool’s ID under conversation_config.agent.prompt.tool_ids. Tools are now decoupled from agents; using tool_ids is the supported way (legacy prompt.tools is deprecated). (elevenlabs.io)

Notes

  • System tools (e.g., language_detection, end_call) don’t use OAuth and are referenced via prompt.built_in_tools, not by ID. (elevenlabs.io)
  • For a concrete OAuth example (Salesforce), see the guide showing how to configure an OAuth2 Workspace Auth Connection and then reference it in webhook tool configurations. (elevenlabs.io)

Key docs:

  • Create tool (with api_schema.auth_connection.auth_connection_id): POST /v1/convai/tools. (elevenlabs.io)
  • Create agent: POST /v1/convai/agents/create. (elevenlabs.io)
  • Tools decoupling and prompt.tool_ids. (elevenlabs.io)
  • Supported tool auth methods and where to configure them. (elevenlabs.io)

Sources:
Server tools | ElevenLabs Documentation
Create tool | ElevenLabs Documentation
Server tools | ElevenLabs Documentation
Create agent | ElevenLabs Documentation
System tools | ElevenLabs Documentation
Salesforce | ElevenLabs Documentation
Create tool | ElevenLabs Documentation
Create agent | ElevenLabs Documentation
Agent tools deprecation | ElevenLabs Documentation
Server tools | ElevenLabs Documentation

thin topaz
# blazing socket

Why setting builtInTools via elevenlabs conversational sdk like endCall or languageDetection doesn’t enable it in the dashboard?