EMAIL & COMMUNICATION · FOR AI AGENTS
LiveKit Cloud
Run realtime voice agents on managed media and deployment infrastructure.
Deploy agent code with the LiveKit CLI, connect participants through authenticated rooms, and use managed realtime transport with your chosen speech and language models.
Your integration checklist
- Capabilities
- What you need from your owner
- Create a LiveKit Cloud project, choose a plan, prepare a working agent and configure its secrets. Link the CLI to the intended project and implement an authenticated token endpoint for production clients.
- Authentication
- CLI setup uses browser sign-in through lk cloud auth. Room participants use JWT access tokens created on a trusted backend with project credentials; the development token server is not a production authentication system.
- Your first task
- Inspect the linked project, deploy a development agent and connect one controlled client. Verify the expected agent joins the intended room and responds.
- Billing & payment
- Cloud plans meter resource consumption. Agent session time, media transport, inference, observability and phone-number usage can contribute separately; estimate the complete session rather than only model tokens.
- Agent payment support
- Not verified. Programmatic usage does not establish autonomous checkout support.
- Operational limits
- Plan quotas and concurrency limits apply. Free Build allowances cap new requests when exhausted. Your application still owns agent behavior, tool permissions and session shutdown.
ORIGINAL AI AGENT STORE INTEGRATION PLAYBOOK
Verify dispatch, media flow and shutdown as separate session states
Editorial recommendation · Reviewed 2026-10-21
When to choose this service
Choose LiveKit Cloud when an agent needs a persistent deployment and realtime audio connection to users, while retaining control over its application code and model choices.
The service supplies operated media transport and agent hosting. Generating voice-agent code does not supply network routing, running containers or capacity for simultaneous sessions.
Suggested workflow
- Resolve the project, agent dispatch name and deployment configuration. Keep development credentials and test rooms separate from production traffic.
- Connect a controlled client with a backend-issued token. Check the selected agent, audio input and output, interruption handling and tool permissions in the same session.
- End the session explicitly and inspect its lifecycle and usage. Test the expected concurrency before enabling wider access.
Acceptance checks before real work
- Confirm the dispatched agent belongs to the intended project and deployment.
- Check audio in both directions rather than treating a joined room as a successful conversation.
- Verify session termination and all relevant usage meters.
If the workflow fails
Inspect room, deployment and session logs before dispatching another agent. Recover the client connection separately from agent execution so reconnecting does not duplicate external actions.