i run the python starter repo on my computer, it uses inference for the models so its livekit inference, if i do lk deploy, then does it go to livekit cloud, I dont what to buy a subscrtiption build, instead i want to pay per model use only thro the free subscription

Thanks for reaching out, Mohammed. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

so if i do lk agent deploy, I wouldnt have to buy subscription and just do pay as you go?

Yes — that’s correct. • Running

lk agent deploy

deploys your agent to LiveKit Cloud, which is required to use LiveKit Inference. • You do not need to buy a paid Ship/Scale subscription to use Inference. The free Build plan includes agent session minutes and some LiveKit Inference credits. • After you consume the included credits, inference is billed pay-as-you-go (for example, per-minute for STT or per-character for TTS) at the standard rates — you’ll only pay for additional model usage. Would you like help: • confirming your project is linked to LiveKit Cloud before you run

lk agent deploy

, or • checking which inference credits or rates apply to the specific models you plan to use?, with interactive elements

do i have limited agent sessions in build plan? doesnt that go as pay as u go?

Thanks — good follow-up. I’ll clarify how agent sessions and billing work on the Build plan. Short answer • The Build plan is free and includes 1,000 agent session minutes per month. • Agent sessions are limited to that included allotment on the Build plan. After you use the included minutes, additional agent session usage is billed pay-as-you-go. • Inference usage (models) also has included credits on Build and is billed separately per model consumption once included credits are exhausted (for example, per-minute for STT or per-character for TTS). What to keep in mind • The free Build plan gives you a monthly allotment (1,000 agent session minutes plus some inference credits). These included minutes are not unlimited. • Once included agent session minutes are consumed, you pay for additional agent session minutes on a pay-as-you-go basis. • Model (inference) billing is separate from agent session minutes and follows pay-as-you-go after any included inference credits are used. Next steps I can help with • I can check how many session minutes you’ve used this month and how many inference credits remain. • I can show how to estimate costs for the models you plan to run (give me which models and expected usage pattern). Which would you like?, with interactive elements

i have build a RAG and added it as a tool, so will that also be deployed in livekit cloud?

Thanks — I’ll clarify this for you. Acknowledging your question • You asked whether your RAG tool (added as a tool to your agent) will be deployed to LiveKit Cloud when you run

lk agent deploy

, and how that affects billing. Short answer • Yes —

lk agent deploy

packages and deploys your entire agent codebase to LiveKit Cloud, including any custom tools like your RAG implementation. • On the free Build plan you get monthly free allowances (for example, 1,000 agent session minutes and some inference credits). Once those are used, additional agent session time and inference/model usage are billed pay-as-you-go. What that means for your RAG tool • Your RAG logic, connectors, and any provider keys/configurations included in the agent project will run in the cloud after deployment. • RAG behavior that triggers model inference (LLM calls, STT/TTS) will consume inference credits or incur pay-as-you-go charges once included credits are exhausted. • Time the agent is active (agent session minutes) counts toward your Build plan allotment; extra minutes are billed at the pay-as-you-go rate. If you want to control costs • Monitor your agent session minutes and inference credit usage. • Use lower-cost models or fewer long-running sessions to reduce inference and session spend. • Optionally bring your own provider API keys for LLM/STT/TTS if you prefer to avoid LiveKit inference charges. Would you like help with any of the following? • Confirming that your project is linked to LiveKit Cloud before you run

lk agent deploy

. • Checking your current Build-plan usage (session minutes and inference credits). • Estimating costs for your expected RAG usage (tell me which models and approximate usage pattern). Sources: examples/drive-thru | livekit/livekit_composite | DeepWiki | Agents (Node.js) | livekit/livekit_composite | DeepWiki | Self-hosted deployments | LiveKit Documentation | Agents (Python) | livekit/livekit_composite | DeepWiki | Google AI and LiveKit | LiveKit Documentation, with interactive elements