Here’s the fastest, doc-backed path to create and deploy a Voice AI agent with ElevenLabs:
Create the agent
Use the Quickstart to create an agent via the dashboard, API, or CLI. The guide also shows testing in the dashboard and a 1‑line web embed snippet. (elevenlabs.io)
Configure behavior and voice
Set “First message” and “System prompt,” then pick a voice in the Voice tab. (elevenlabs.io)
Attach documents/URLs in the agent’s Knowledge Base; manage and reuse docs from the Knowledge Base dashboard. (elevenlabs.io)
Add tools (actions)
Enable Client Tools, Server/Webhook Tools, MCP Tools, and use built‑in System Tools to let the agent take actions during conversations. (elevenlabs.io)
Secure access
Use signed URLs (recommended for client apps) or host allowlists; signed URLs are generated server‑side and expire after 15 minutes. (elevenlabs.io)
Embed or connect in real time
Easiest: paste the agent widget snippet from Quickstart into your site. (elevenlabs.io)
Programmatic: connect over the Agents WebSocket API (wss://api.elevenlabs.io/v1/convai/conversation?agent_id=...). Docs include auth via signed URLs and event formats. (elevenlabs.io)
Telephony (phone calls)
Twilio native integration supports inbound and outbound; import your Twilio number and assign it to the agent. (elevenlabs.io)
Or connect any SIP‑compatible system via SIP trunking; use sip.rtc.elevenlabs.io and assign the imported number to your agent. (elevenlabs.io)
Monitor and iterate
Use Analytics, Call History, experiments, and analysis tooling from the dashboard/docs to evaluate and improve agents. (elevenlabs.io)
Limits and pricing to plan deployments
Concurrency (parallel calls) depends on plan; current per‑workspace limits are listed here. (help.elevenlabs.io)
gusty crypt
fervent current