How do you deploy a Langflow agent in 2026?

As of April 2026, Langflow agents are deployed by exporting the flow as JSON, running Langflow in production via Docker or DataStax-managed hosting, and exposing the flow as a REST endpoint at `/api/v1/run/ `. Authentication uses API keys configured per workspace.

Deploy a Langflow Agent (2026)

Deploying a Langflow Agent

Langflow exposes any flow as a REST API once Langflow is running in a server context. As of April 2026, three deployment patterns dominate.

Step 1 — Build and Export the Flow

In the Langflow UI, build the agent flow (chain, agent, memory, tools). Click the Export button to download the flow as JSON. The JSON includes node configuration, edges, and component versions, making it portable across instances.

Step 2 — Choose a Hosting Option

Three common options:

Docker self-host — docker run -p 7860:7860 langflowai/langflow:latest runs Langflow on a single host. Mount a volume for persistent flows and use Postgres for the metadata database in production.
Kubernetes — The official Helm chart deploys Langflow with autoscaling and Postgres. Suits production deployments above ~10 concurrent users.
DataStax-managed — DataStax operates a managed Langflow service with Astra DB integration for vector storage. Pricing follows Astra DB consumption.

Step 3 — Expose the Flow as an API

Once Langflow is running, every saved flow is accessible at:

curl -X POST http://your-host/api/v1/run/<flow_id>?stream=false \
  -H "Content-Type: application/json" \
  -H "x-api-key: $LANGFLOW_API_KEY" \
  -d '{"input_value": "hello", "output_type": "chat"}'

Generate API keys per workspace from Settings → API Keys.

Step 4 — Add Observability

Enable LangSmith or OpenTelemetry exporters in langflow.yaml to trace agent steps, token usage, and tool calls. For framework-level monitoring, pair Langflow with LangChain callbacks.

Step 5 — Version and Promote

Treat exported JSON like code: commit to git, run integration tests against staging Langflow, then import to production. Langflow flows can be diffed and reviewed in pull requests.