Thanks — I understand you’re seeing agents in LiveKit Cloud but some incoming calls aren’t being picked up. Below is a focused troubleshooting checklist you can run through to find and fix the most common causes. Quick checklist (start here) • Confirm you have multiple worker processes for each agent. Run

lk agent status

to verify replica counts. • Check agent logs in the dashboard for errors at the time of missed calls. • Verify dispatch rules are routing the call to the expected agent(s) and trunk IDs are correct. • Ensure your agent hosts aren’t CPU/memory/network constrained when calls arrive. Step-by-step troubleshooting 1. Verify agent availability and replicas • Ensure at least 3 worker instances per agent so jobs fail over and concurrency is handled. Low replica counts cause intermittently missed jobs. • Use

lk agent status

to confirm replica counts and whether workers are “ready.” 1. Inspect logs and metrics • Look at agent logs for errors when a call is missed: connection timeouts, job assignment failures, or room creation delays. • Check agent metrics (connection latency, job queue length) around the missed-call timestamps to spot bottlenecks. 1. Confirm dispatch rule configuration • Check the SIP dispatch rule used for these calls. Ensure it: ◦ Targets the correct agent(s). ◦ Has the right

trunk_ids

associated. ◦ Uses the intended dispatch type (for example, individual rooms per call if that’s required). • Misconfigured dispatch rules often route calls to the wrong agent group or no agent at all. 1. Check resource contention and scaling • Make sure no other processes are consuming heavy CPU/memory/network on agent hosts when calls arrive. • If you self-host: scale horizontally earlier than the agent’s default

load_threshold

(e.g., scale when CPU ~50% rather than 70%). • If using LiveKit Cloud: confirm your setup can scale quickly enough for your peak call bursts. 1. Confirm room creation/connect timing • Look for errors like “room connection was not established within X seconds.” If room setup is slow, agents may time out before joining. • If necessary, increase timeouts or optimize the steps that run when a call is created (fewer synchronous tasks before agent assignment). 1. Reproduce in a controlled test • Simulate incoming calls in a test environment or using an evaluation agent to observe the routing and assignment behavior. • Run repeated tests while increasing load to see at what point assignments fail. If you want, I can help further — please share: • The output of

lk agent status

(or a screenshot). • Example timestamps and call IDs for a few missed calls. • The dispatch rule settings (name and trunk_ids) for the calls that weren’t picked up. With those details I can point to likely root causes (replica count, dispatch rule mismatch, resource exhaustion, or timing) and recommend next actions tailored to your setup. Sources: Agents Framework | livekit/livekit_composite | DeepWiki | examples/frontdesk | livekit/livekit_composite | DeepWiki | examples/drive-thru | livekit/livekit_composite | DeepWiki | Self-hosted deployments | LiveKit Documentation | Build Your First AI Voice Agent in Python: Complete Tutorial | LiveKit