Deploy flows on watsonx Orchestrate

import Icon from "@site/src/components/icon"; import Tabs from '@theme/Tabs'; import TabItem from '@theme/TabItem';

:::tip As of Langflow 1.10.x, the IBM watsonx Orchestrate deployments feature is behind a feature flag. To enable it, set the following environment variable before starting Langflow:

LANGFLOW_FEATURE_WXO_DEPLOYMENTS=true

:::

Create a flow and deploy it to IBM watsonx Orchestrate.

Deploying a flow on IBM watsonx Orchestrate is different from the other Langflow deployment options. This workflow does not deploy a full-featured Langflow server and flow builder UI. Instead, Langflow packages your selected flow and flow version, and then publishes it to IBM watsonx Orchestrate as a wxO Tool that a wxO Agent can execute. Langflow is used to build and configure the flow, while IBM watsonx Orchestrate hosts the wxO Agent experience and invokes the deployed flow as part of that wxO Agent's toolset.

Prerequisites

• Install and start Langflow
• Create an OpenAI API key
• Create an IBM watsonx Orchestrate instance

Create and deploy a flow

1. Create a flow in the Langflow UI, such as the Simple Agent starter flow in the Quickstart.

2. To deploy this flow to watsonx Orchestrate, click <Icon name="Rocket" aria-hidden="true"/> Deploy. The Provider pane opens.

3. Enter the Name, Service Instance URL, and API Key from your IBM watsonx Orchestrate instance. These values are found in the Settings page of your IBM watsonx Orchestrate instance.

   • Name: YOUR_DEPLOYMENT_NAME
   • Service Instance URL: https://api.dl.watson-orchestrate.ibm.com/instances/80194572-4421-6735-91ab-55c0d8e4f962
   • API Key: YOUR_WATSONX_ORCHESTRATE_API_KEY

   The last segment of the Service Instance URL is the IBM watsonx Orchestrate tenant ID, which can be found in your watsonx Orchestrate deployment. In this example, the tenant ID is 80194572-4421-6735-91ab-55c0d8e4f962.

4. Click Next. The Deployment Type pane opens.

5. Enter a Type, Name, Model, and Description for your wxO Agent, and then click Next.

   The Type is always Agent. The deployed flow is a wxO Agent with your flow available as a wxO Tool the wxO Agent can call.

   The Model list is populated from the connected watsonx Orchestrate instance, not Langflow.

6. In the Flows pane, in the Available list, select a flow and flow version to deploy. For this example, select the simple agent flow you created. By default, the flow is version 1.

   Optionally, to attach an additional version, select the same flow again from the list and choose a different version. You can attach multiple versions of the same flow to a single wxO Agent. Each flow version is deployed as a separate wxO Tool with a unique name to prevent naming collisions in watsonx Orchestrate.

7. To configure the watsonx Orchestrate connection, click Create Connection. In this tab, you can create a new connection, or select an existing connection to bind to the flow.

   To create a new connection, do the following:

   1. Enter a Connection Name and any environment variables the flow requires, such as the OPENAI_API_KEY for the simple agent flow. Langflow auto-detects global variables from the flow JSON file. Click <Icon name="Plus" aria-hidden="true"/> Add variable to add any additional environment variables.

   2. To add the new connection to the list of available connections, click Create Connection.

   3. In the list of available connections, select the new connection, and then click Attach Connection to Flow.

   :::tip To bind the connection to the flow without environment variable binding, click Skip, and then click Next. :::

8. Click Next. The Review & Confirm pane opens.

9. Confirm the deployment values are correct, and then click <Icon name="Rocket" aria-hidden="true"/> Deploy.

   Langflow installs any required extra dependencies on your watsonx Orchestrate tenant automatically.

   In the Langflow UI, Deployment successful indicates your deployment succeeded.

   :::tip If you get an error that the wxO Tool name already exists on your deployment, click <Icon name="Pencil" aria-hidden="true"/> Edit to change the wxO Tool name. :::

10. Click Test to open a chat window with your wxO Agent on watsonx Orchestrate. Enter a question, and the wxO Agent responds using the connected flow as a wxO Tool.

11. In your IBM watsonx Orchestrate deployment, navigate to Agent chat > Manage Agents, and confirm that a new wxO Agent is listed with your Langflow flow attached as a wxO Tool.

12. To test the wxO Agent in watsonx Orchestrate, click Talk to agent, and then enter a question. The wxO Agent responds using your Langflow flow as a wxO Tool.

13. To deploy the wxO Agent to a live environment, click Deploy.

14. Optionally, click Activate agent monitoring.

15. To view your wxO Agent's metrics, click Agent chat > Agent analytics, and then select your wxO Agent. This page includes metrics for your wxO Agent, such as Total messages, Failed messages, and Latency average. To view traces for each wxO Agent request, click Trace Detail. For more information, see Monitoring agents.

Manage deployments in Langflow

From the Projects page, click Deployments to open the deployment management screen.

• Deployments:

  A Deployment is a published wxO Agent equipped with one or more Langflow flows as wxO Tools. Deployment details include the wxO Agent name, type, attached flows, model, and the IBM watsonx Orchestrate environment it belongs to.

  Use the Deployments tab to create, update, view, and delete flow deployments in Langflow.

• Deployment Environments:

  A Deployment Environment is a saved watsonx Orchestrate target that Langflow can deploy to. An environment stores the connection details for a watsonx Orchestrate tenant.

  Use the Deployment Environments tab to connect, view, and disconnect IBM watsonx Orchestrate environments in Langflow.

  To manage the tenant itself, use the IBM watsonx Orchestrate dashboard.

Send requests to your flow

After you deploy your flow to IBM watsonx Orchestrate, you can chat with the wxO Agent through the Langflow deployment run endpoints.

Don't use the /run endpoint for flows deployed to IBM watsonx Orchestrate. Instead use POST /api/v1/deployments/{deployment_id}/runs to start a run, and GET /api/v1/deployments/{deployment_id}/runs/{run_id} to check its status.

Endpoint paths must be prefixed with your Langflow server URL, such as http://localhost:7860.

To find your deployment_id, navigate to the Projects page, click Deployments, and select the deployment. The ID is displayed in the deployment details.

These endpoints require a Langflow API key passed in the x-api-key header. To create one, see API keys and authentication.

Create deployment run endpoint

Endpoint: POST /api/v1/deployments/{deployment_id}/runs

Description: Start a run for a deployed wxO Agent and return a provider-owned run ID that you can poll for status.

Example request

import requests

url = "http://LANGFLOW_SERVER_ADDRESS/api/v1/deployments/DEPLOYMENT_ID/runs"

payload = {
    "provider_data": {
        "input": "Summarize today's tickets",
        "thread_id": "thread-123"
    }
}

headers = {
    "Content-Type": "application/json",
    "x-api-key": "LANGFLOW_API_KEY"
}

response = requests.post(url, json=payload, headers=headers)
response.raise_for_status()

print(response.json())

const payload = {
  provider_data: {
    input: "Summarize today's tickets",
    thread_id: "thread-123"
  }
};

const options = {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "x-api-key": "LANGFLOW_API_KEY"
  },
  body: JSON.stringify(payload)
};

fetch("http://LANGFLOW_SERVER_ADDRESS/api/v1/deployments/DEPLOYMENT_ID/runs", options)
  .then((response) => response.json())
  .then((response) => console.log(response))
  .catch((err) => console.error(err));

curl --request POST \
     --url "http://LANGFLOW_SERVER_ADDRESS/api/v1/deployments/DEPLOYMENT_ID/runs" \
     --header "Content-Type: application/json" \
     --header "x-api-key: LANGFLOW_API_KEY" \
     --data '{
       "provider_data": {
         "input": "Summarize today's tickets",
         "thread_id": "thread-123"
       }
     }'

Request body

Example response

{
  "deployment_id": "3ea34379-1f72-4a33-9f6e-9e3ca88365b5",
  "provider_data": {
    "id": "run-42",
    "agent_id": "agent-123",
    "thread_id": "thread-123",
    "status": "accepted",
    "result": null,
    "started_at": null,
    "completed_at": null,
    "failed_at": null,
    "cancelled_at": null,
    "last_error": null
  }
}

Response body

The response returns the Langflow deployment_id and a provider_data object containing the provider-owned run metadata. Use provider_data.id as the run_id when checking the run status.

Get deployment run status endpoint

Endpoint: GET /api/v1/deployments/{deployment_id}/runs/{run_id}

Description: Retrieve the current status and result of a deployment run.

Example request

import requests

url = "http://LANGFLOW_SERVER_ADDRESS/api/v1/deployments/DEPLOYMENT_ID/runs/RUN_ID"

headers = {
    "Content-Type": "application/json",
    "x-api-key": "LANGFLOW_API_KEY"
}

response = requests.get(url, headers=headers)
response.raise_for_status()

print(response.json())

const options = {
  method: "GET",
  headers: {
    "Content-Type": "application/json",
    "x-api-key": "LANGFLOW_API_KEY"
  }
};

fetch("http://LANGFLOW_SERVER_ADDRESS/api/v1/deployments/DEPLOYMENT_ID/runs/RUN_ID", options)
  .then((response) => response.json())
  .then((response) => console.log(response))
  .catch((err) => console.error(err));

curl --request GET \
     --url "http://LANGFLOW_SERVER_ADDRESS/api/v1/deployments/DEPLOYMENT_ID/runs/RUN_ID" \
     --header "Content-Type: application/json" \
     --header "x-api-key: LANGFLOW_API_KEY"

Path parameters

Example response

{
  "deployment_id": "3ea34379-1f72-4a33-9f6e-9e3ca88365b5",
  "provider_data": {
    "id": "run-42",
    "agent_id": "agent-123",
    "thread_id": "thread-123",
    "status": "completed",
    "result": {
      "output": "Here is your summary..."
    },
    "started_at": "2026-04-03T12:40:00Z",
    "completed_at": "2026-04-03T12:40:05Z",
    "failed_at": null,
    "cancelled_at": null,
    "last_error": null
  }
}

Response body

Check provider_data.status to determine whether the run is still processing or has finished. When the status is completed, read the output from provider_data.result.

If provider_data.status is failed, the provider_data.failed_at timestamp and provider_data.last_error field contain details about what went wrong.