Skip to main content
Strands Agents is a framework for building AI agents with a focus on simplicity and performance. For more details on Strands Agents, refer to the official Strands Agents documentation. LangWatch can capture traces generated by Strands Agents by leveraging its built-in OpenTelemetry support. This guide will show you how to set it up.

​
Prerequisites

  1. Install LangWatch SDK and Strands Agents: 
    pip install langwatch strands-agents[otel] strands-agents-tools
  2. Set up your LLM provider: You’ll need to configure your preferred LLM provider (OpenAI, Anthropic, AWS Bedrock, etc.) with the appropriate API keys.

​
OpenTelemetry Setup Options

LangWatch supports three approaches for instrumenting Strands Agents with OpenTelemetry: This is the simplest approach where LangWatch handles all OpenTelemetry setup: 
import langwatch

# Initialize LangWatch - API key is set from environment variable automatically
langwatch.setup()

​
Option 2: StrandsTelemetry with Custom Configuration

For more control over OpenTelemetry configuration, you can use StrandsTelemetry: 
import os
from strands import Agent
from strands.telemetry import StrandsTelemetry
import langwatch

# Configure StrandsTelemetry with LangWatch endpoint
strands_telemetry = StrandsTelemetry()
strands_telemetry.setup_otlp_exporter(
    endpoint=f"{os.environ.get('LANGWATCH_ENDPOINT', 'https://app.langwatch.ai')}/api/otel/v1/traces",
    headers={"Authorization": "Bearer " + os.environ["LANGWATCH_API_KEY"]},
)

# Skip LangWatch OpenTelemetry setup since StrandsTelemetry handles it
langwatch.setup(skip_open_telemetry_setup=True)

​
Option 3: Skip OpenTelemetry Setup (When Already Configured)

If OpenTelemetry is already configured by another component in your application (like FastAPI, Django, or another framework), you can skip LangWatch’s OpenTelemetry setup: 
import langwatch

# Skip OpenTelemetry setup since it's already configured elsewhere
langwatch.setup(skip_open_telemetry_setup=True)
This is useful when:
  • Your backend or infrastructure framework already sets up OpenTelemetry
  • You have a custom OpenTelemetry configuration
  • Multiple components in your stack configure OpenTelemetry

​
Basic Agent Setup

Here’s a complete example showing how to create and instrument a Strands Agent: 
from strands import Agent
from strands.models.litellm import LiteLLMModel
import langwatch

# Initialize LangWatch
langwatch.setup()

class MyAgent:
    def __init__(self):
        # Configure the model using LiteLLM for provider flexibility
        self.model = LiteLLMModel(
            client_args={
                "api_key": os.getenv("OPENAI_API_KEY"),
            },
            model_id="openai/gpt-5-mini",
        )
        
        # Create the agent with tracing attributes
        self.agent = Agent(
            name="my-agent",
            model=self.model,
            system_prompt="You are a helpful AI assistant.",
            tools=[],  # Add your tools here
            trace_attributes={
                "custom.model_id": "openai/gpt-5-mini",
                "custom.example.attribute": "swift",
            },
        )

    def run(self, prompt: str):
        return self.agent(prompt)

# Use the agent
agent = MyAgent()
response = agent.run("Hello, how can you help me?")
print(response)

​
Integration with Web Frameworks

​
Chainlit Integration

Here’s how to integrate Strands Agents with Chainlit while maintaining full observability: 
import os
from strands import Agent
from strands.models.litellm import LiteLLMModel
import langwatch
import chainlit.config as cl_config
import chainlit as cl
from dotenv import load_dotenv

load_dotenv()

# Chainlit has broken telemetry, so we need to disable it
cl_config.config.project.enable_telemetry = False

# Initialize LangWatch
langwatch.setup()

class KiteAgent:
    def __init__(self):
        self.model = LiteLLMModel(
            client_args={
                "api_key": os.getenv("OPENAI_API_KEY"),
            },
            model_id="openai/gpt-5-mini",
        )
        self.agent = Agent(
            name="kite-agent",
            model=self.model,
            system_prompt="You are a helpful AI assistant.",
            tools=[],
            trace_attributes={
                "custom.model_id": "openai/gpt-5-mini",
                "custom.example.attribute": "swift",
            },
        )

    def run(self, prompt: str):
        return self.agent(prompt)

@cl.on_message
@langwatch.trace()
async def main(message: cl.Message):
    msg = cl.Message(content="")

    # Update the current trace with additional metadata
    langwatch.get_current_trace().update(
        metadata={
            "custom.example.attribute": "swift",
        }
    )

    agent = KiteAgent()
    response = agent.run(message.content)

    await msg.stream_token(str(response))
    await msg.update()

​
Adding Custom Attributes and Metadata

You can add custom attributes to your traces in several ways:

​
Agent-Level Attributes

agent = Agent(
    name="my-agent",
    model=model,
    system_prompt="You are a helpful AI assistant.",
    trace_attributes={
        "custom.model_id": "openai/gpt-5-mini",
        "custom.environment": "production",
        "custom.service": "customer-support",
    },
)

​
Function-Level Metadata

@langwatch.trace(name="Handle Request")
def handle_request(input_message: str):
    langwatch.get_current_trace().update(
        metadata={
            "user_id": "user_001",
            "thread_id": "thead_001"
        }
    )

    return my_custom_strands_agents_app(input_message)

@langwatch.span(name="Handle")
def my_custom_strands_agents_app(input_message: str):
    # Update the current trace with additional metadata
    current_trace = langwatch.get_current_trace()
    if current_trace:
        current_trace
    
    response = agent(input_message)
    return response

​
How it Works

  1. LangWatch Setup: langwatch.setup() initializes the LangWatch SDK and sets up OpenTelemetry tracing.
  2. Model Configuration: Use LiteLLMModel for flexible provider support or specific model classes like BedrockModel for AWS Bedrock.
  3. Agent Creation: The Agent constructor accepts trace_attributes for consistent metadata across all traces.
  4. Automatic Tracing: All agent interactions, model calls, and tool executions are automatically traced and sent to LangWatch.
  5. Custom Metadata: Use @langwatch.trace() decorators and langwatch.get_current_trace().update() to add context-specific metadata.

​
Environment Variables

Set up your environment variables in a .env file: 
LANGWATCH_API_KEY=your-langwatch-api-key
OPENAI_API_KEY=your-openai-api-key
LANGWATCH_ENDPOINT=https://app.langwatch.ai  # Optional, defaults to this value

​
Notes

  • The strands-agents[otel] package includes OpenTelemetry support out of the box.
  • The trace_attributes parameter allows you to add consistent metadata to all traces from a specific agent instance.
  • For advanced configuration, see the Python integration guide.

​
Troubleshooting

  • Make sure your LANGWATCH_API_KEY is set in the environment.
  • If you see no traces in LangWatch, check that the telemetry is properly configured and that your agent code is being executed.
  • If you aren’t using the LangWatch SDK’s automatic OpenTelemetry setup and traces are not showing, double check the url and path given to OpenTelemetry.

​
Next Steps

Once you have instrumented your code, you can manage, evaluate and debug your application:
  • View traces in the LangWatch dashboard
  • Add evaluation scores to your traces
  • Create custom dashboards for monitoring
  • Set up alerts for performance issues
  • Export data for further analysis

​
Learn More

For more detailed information, refer to the official documentation and other examples:
⌘I