> For the complete documentation index, see [llms.txt](https://docs.roboflow.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.roboflow.com/deployment/roboflow-cloud/cloud-streaming.md). # Cloud Streaming Connect live cameras, run Workflows on managed cloud GPUs, and consume video and prediction results with Cloud Streaming. Cloud Streaming runs a Workflow on a live camera using Roboflow-managed compute. Use it for continuous RTSP, IP, and USB camera processing without hosting an Inference Server. Connect a source, test a Workflow, then choose when to deploy and stop it. Start with the [Cloud Streaming quickstart](/deployment/roboflow-cloud/cloud-streaming/quickstart.md). ## About Cloud Streaming uses three parts: * A source supplies video. It can be a camera connected through the Roboflow Video Connector, a direct RTMP publisher, or a temporary browser webcam or phone session. * A connector runs on a computer near your cameras. It discovers local sources and sends video to Roboflow when requested. It does not run the Workflow's model inference. * A job runs one Workflow on one source. A preview is a bounded test run. A deployment keeps processing independently of your browser until stopped, subject to source availability, limits, and recovery policy. A source can have a deployed Workflow and a separate draft preview. A source being online does not mean a Workflow is processing it. Closing a video player does not stop a deployment.
TaskUse
Run a live camera on managed cloud computeCloud Streaming
Try your webcam or phone cameraA temporary Cloud Streaming preview
Process stored videos or images to completionBatch Processing
Send individual images or requests to a WorkflowServerless Cloud API
Keep processing on your hardware or local networkSelf-hosted Inference Server
The older [Serverless Video Streaming API](/deployment/roboflow-cloud/serverless-api/serverless-video-streaming-api.md) is deprecated. Cloud Streaming is the recommended managed live-camera path. It uses source and job APIs, not a different URL for the older WebRTC client. Self-hosted WebRTC remains a separate supported deployment path. {% hint style="info" %} Cloud Streaming is publicly available in all environments. {% endhint %} ## Web App Open "Deployments", then "Cloud Streaming" in your workspace. You need permission to view streams, and source-management permission to connect cameras or start and stop processing. If access is denied, ask your workspace administrator to check your permissions. Follow these guides in order: 1. [Install a connector](/deployment/roboflow-cloud/cloud-streaming/install-a-connector.md) on the camera's computer or network. 2. [Connect a source](/deployment/roboflow-cloud/cloud-streaming/connect-a-source.md), including RTSP, USB, ONVIF, and temporary cameras. 3. [Run a Workflow](/deployment/roboflow-cloud/cloud-streaming/run-a-workflow.md) and understand preview, deployment, and version behavior. 4. [Consume results](/deployment/roboflow-cloud/cloud-streaming/consume-results.md) as video, JSON, Workflow sink events, or saved captures. 5. [Manage and monitor](/deployment/roboflow-cloud/cloud-streaming/manage-and-monitor.md) your running deployments. Processing uses credits, including previews. Review the displayed estimate before starting and see [pricing and limits](/deployment/roboflow-cloud/cloud-streaming/pricing-and-limits.md). Stop processing explicitly when you finish. ## Roboflow Agent Ask the [Roboflow Agent](https://docs.roboflow.com/agents/roboflow-agent) to connect a camera and preview your Workflow. The Agent can help select a source, inspect results, and deploy when you explicitly ask. A webcam or phone QR session is temporary and cannot become a continuous deployment. Use the [Agent quickstart](/deployment/roboflow-cloud/cloud-streaming/quickstart.md#roboflow-agent) for a complete connect, preview, deploy, and stop sequence. ## HTTP API The public API supports source discovery, Workflow interface inspection, source previews, temporary sessions, and job creation, status, cancellation, and output access. Start with [source setup](/deployment/roboflow-cloud/cloud-streaming/connect-a-source.md#http-api), then [job control](/deployment/roboflow-cloud/cloud-streaming/run-a-workflow.md#http-api). Connector administration, direct-source creation and key rotation, restart, and preview promotion are app operations. Do not substitute private app routes for a public API. The public job-create API also does not provide the app's automatic resource-sizing flow. ## Add a Video Source Open "Deployments", then "Cloud Streaming", and click "Add Video Source". Follow [Connect a Source](/deployment/roboflow-cloud/cloud-streaming/connect-a-source.md) to connect your camera. ## Add Network Cameras From Roboflow Open the connector page and use "RTSP Cameras" to add an RTSP URL or an ONVIF camera found on the network. See [app-managed cameras](/deployment/roboflow-cloud/cloud-streaming/connect-a-source.md) for credentials, profile selection and connector requirements. ## Run a Workflow Select a source and Workflow, then click "Start Processing". See [Run a Workflow](/deployment/roboflow-cloud/cloud-streaming/run-a-workflow.md) for inputs, resource sizing, updates and stopping. ## Deploy a Preview Stream Use "Deploy Stream" on an eligible preview to keep its exact Workflow and settings running. See [preview deployment](/deployment/roboflow-cloud/cloud-streaming/run-a-workflow.md) for source eligibility, sink settings and version behavior. ## Pricing Processing uses credits, including previews. See [Pricing and Limits](/deployment/roboflow-cloud/cloud-streaming/pricing-and-limits.md) for estimates, resource adjustments and credit limits.