can I build my own livekit agent image and run that using LK agent cloud ?

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

can I build the image myself ?

You can build the agent image yourself. LiveKit Cloud Agents supports using a custom container image (your Dockerfile and build artifacts) instead of requiring the platform to build it for you. A few practical notes to help you: • Provide a Dockerfile: Include a standard Dockerfile in your agent project that installs your dependencies and sets the correct entrypoint for your agent process. The CLI will generate a Dockerfile if you want a starting point, but you can replace it with your own. • livekit.toml / registration: Your project should still include the agent configuration (for example

livekit.toml

) so the platform can register and deploy the image with your Cloud project. • Building & pushing: You can build and push the image to a container registry you control (e.g., Docker Hub, ECR, GCR). Ensure the registry is accessible to LiveKit Cloud or use the deployment flow that references your registry credentials. • Deploying with the CLI: You can still use

lk agent create

to register and deploy the agent; the platform will accept your custom image or Dockerfile during that process. • What LiveKit Cloud still provides: Once deployed, the Cloud handles provisioning, health checks, load balancing, logging, rollbacks, and scaling for your agent instances. If you want, I can: • Review a Dockerfile you plan to use (paste it here), • Show an example Dockerfile layout for an agent, • Walk through the steps to push an image to a registry and reference it from the LiveKit deployment. Which would you like next?

Walk through the steps to push an image to a registry and reference it from the LiveKit deployment.

show me the command to reference a image

Thanks — I’m replying to your request for the exact command to reference a custom agent image and how to push it to a registry for use with LiveKit. Short answer • To push an image to a registry: ◦ docker build -t your-registry/your-agent-image:tag . ◦ docker push your-registry/your-agent-image:tag • To run that image so it registers with a LiveKit server (self-hosted agent), run: ◦ docker run -e LIVEKIT_URL=<your-livekit-url> -e LIVEKIT_API_KEY=<your-key> -e LIVEKIT_API_SECRET=<your-secret> your-registry/your-agent-image:tag Important limitation • LiveKit Cloud’s agent creation flow builds the image for you from source; you cannot make the Cloud use an arbitrary external image during

lk agent create

. If you want to use a pre-built image you control, run the agent on your own infrastructure (Docker, Kubernetes, VMs) so the agent process can connect to LiveKit. See Self-hosted deployments: https://docs.livekit.io/agents/ops/deployment/custom/ Step-by-step examples 1) Build and push (example using Docker Hub or other registry)

# Build
docker build -t <http://docker.io/your-username/my-agent:1.0|docker.io/your-username/my-agent:1.0> .

# Push
docker push <http://docker.io/your-username/my-agent:1.0|docker.io/your-username/my-agent:1.0>

Replace

<http://docker.io/your-username/my-agent:1.0|docker.io/your-username/my-agent:1.0>

with your registry and tag (ECR, GCR, etc. are similar). 2) Run the agent container (registers with LiveKit)

docker run \
  -e LIVEKIT_URL="<wss://your-livekit.example>" \
  -e LIVEKIT_API_KEY="your-key" \
  -e LIVEKIT_API_SECRET="your-secret" \
  <http://docker.io/your-username/my-agent:1.0|docker.io/your-username/my-agent:1.0>

Make sure your image starts the AgentServer process as the container entrypoint and uses those env vars to connect. 3) Kubernetes deployment (example snippet)

apiVersion: apps/v1
kind: Deployment
metadata:
  name: livekit-agent
spec:
  replicas: 2
  selector:
    matchLabels:
      app: livekit-agent
  template:
    metadata:
      labels:
        app: livekit-agent
    spec:
      containers:
      - name: agent
        image: <http://docker.io/your-username/my-agent:1.0|docker.io/your-username/my-agent:1.0>
        env:
        - name: LIVEKIT_URL
          value: "<wss://your-livekit.example>"
        - name: LIVEKIT_API_KEY
          value: "your-key"
        - name: LIVEKIT_API_SECRET
          value: "your-secret"

If you want, I can: • Review your Dockerfile to confirm the agent entrypoint and env var handling. • Show commands for authenticating and pushing to a specific registry (ECR, GCR, ACR). • Help convert your run example into a Helm/Kustomize snippet. Which of those would you like next? Sources: Self-hosted deployments | LiveKit Documentation