For AI agents: a documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.
Call queueing and hold audio: Added call queueing for agents at their concurrency limit. Queued callers hear hold audio and receive queue_status events until they are admitted or time out. New API methods upload or delete custom hold audio.
Test version metadata: Agent test results now identify the agent version they ran against and whether they included draft changes. Invocation summaries also indicate when partial resubmissions ran against different versions.
Model and account metadata: Agent LLM configuration now accepts gemini-3.8-flash. WhatsApp account responses identify whether the account uses the Cloud API or coexistence signup flow.
Music
Music v2.5 API support: Music endpoints and SDKs now accept music_v2_5. Composition plan generation chunks support up to 6,132 characters, with up to 30 lines of 200 characters each.
SDK releases
JavaScript SDK
v2.68.0 - Added methods to upload and delete agent hold audio. Regenerated types for call queueing, procedure references, draft branch creation, RAG query limits, test version metadata, gemini-3.8-flash and music_v2_5.
Python SDK
v2.68.0 - Added synchronous and asynchronous methods to upload and delete agent hold audio. Regenerated types for call queueing, procedures, draft branch creation, RAG query limits, test version metadata, gemini-3.8-flash and music_v2_5. The compose_detailed wrapper now accepts every generated music model ID and both prompt- and chunk-based composition plans.
iOS SDK
v3.3.1 - Reconciles agent messages by response_id instead of event_id, preserving multiple responses around tool calls and preventing streamed parts or corrections from attaching to the wrong message.
ElevenLabs CLI
v1.3.0 - Added optional --intent metadata to generated API commands, with ELEVENLABS_AGENT_INTENT as a task-wide default. Added elevenlabs feedback missing-capability to report workflows that the CLI does not support.
Packages
@elevenlabs/convai-widget-core@0.18.2 and @elevenlabs/convai-widget-embed@0.18.2 - Fixed duplicate agent replies after tool and procedure calls. Streamed text replies retain Markdown formatting, and voice conversations no longer render unspoken streamed chat parts. The embed package also limits user_activity events to once per second and clears the limit when a session ends.
Added optional platform_settings.queueing_config (AgentQueueingConfig) with enabled (boolean, default false) and wait_timeout_seconds (integer, 1–1,800, default 180).
Queue configuration responses include read-only nullable hold_audio (AgentHoldAudioConfig). Agent responses also include optional default_hold_audio_url (string).
Get conversation - GET /v1/convai/conversations/{conversation_id}
Metadata adds optional nullable queue_wait_secs (number). Queue time is excluded from call_duration_secs and billed time.
Test suite invocation and unit test run responses add optional nullable version_id (string) and optional ran_against_draft (boolean, default false).
Invocation summaries also add optional runs_diverged_from_version (boolean, default false).
POST /v1/convai/agents/{agent_id}/knowledge-base/rag-query
The request body adds optional nullable max_documents_length (integer, 1–50,000) and max_retrieved_rag_chunks_count (integer, 1–20) query-specific overrides.
MusicModelID adds music_v2_5 for POST /v1/music/video-to-music, POST /v1/music/plan, Compose music, POST /v1/music/detailed, POST /v1/music/detailed/stream, POST /v1/music/stream, POST /v1/music/upload and POST /v1/music/finetunes.
GenerationChunk.text increases maxLength from 6,000 to 6,132 characters for POST /v1/music/plan, POST /v1/music, POST /v1/music/detailed, POST /v1/music/detailed/stream, POST /v1/music/stream and POST /v1/music/upload. Section names can contain 1–100 characters, and chunks can contain up to 30 lines with 200 characters per line.
Schema changes
ElevenAgents
Real-time server events add queue_status. Its required queue_status_event.status accepts waiting, admitted or timed_out; timed_out precedes WebSocket close code 4300. Hold audio continues to arrive as regular audio events.
The LLM enum adds gemini-3.8-flash, including conversation initiation overrides.
Conversation system tool result unions add start_procedure_success, start_procedure_error, end_procedure_success and end_procedure_error.
FeaturesUsageCommonModel adds optional freeform_procedure and structured_procedure feature statuses.