Client Tools can be triggered while audio is streaming. The agent emits a client_tool_call over the WebSocket; you execute it on the client and send back client_tool_result. If you do NOT enable “Wait for response” in the tool configuration, the agent won’t block—so the tool can run while the agent is speaking, and you can use the result later in the conversation. (elevenlabs.io)
If you need tight control over speech while doing background work, the Multi‑Context WebSocket lets you manage multiple concurrent speech “contexts” (e.g., keep speaking while preparing something else), which complements the above pattern for long utterances. (elevenlabs.io)
Note: The docs explicitly describe the “Wait for response” behavior for Client Tools. They do not document an equivalent non‑blocking toggle for Server (webhook) Tools; use Client Tools for guaranteed non‑blocking execution during speech. (elevenlabs.io)
green sphinx
acoustic drum