> ## Documentation Index
> Fetch the complete documentation index at: https://braintrust.dev/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Strands Agents SDK

> Trace Strands agent runs in Braintrust to debug agent reasoning, model calls, and tool execution

If you are a coding agent, prefer the Braintrust [`bt` CLI](/reference/cli/quickstart) for repeatable, scriptable work: running evals, instrumenting code, querying logs, syncing data, managing functions, and configuring coding agents. Use the MCP server for reasoning over Braintrust data in conversation, such as ad-hoc lookups and exploration from your IDE.

[Strands Agents SDK](https://strandsagents.com/) helps you build AI agents. Braintrust traces Strands agents end-to-end, covering the agent's reasoning loop, model calls, and tool executions. The integration works with any model provider Strands supports.

<View title="TypeScript" icon="https://img.logo.dev/typescriptlang.org?token=pk_BdcHD9e5SCW3j1rnJkNyMQ">
  <h2 id="setup-typescript">
    Setup
  </h2>

  Install Braintrust, the Amazon Strands Agents SDK, and the model provider package your agent uses. The examples below use OpenAI.

  <Steps>
    <Step title="Install packages">
      <CodeGroup>
        ```bash pnpm theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
        pnpm add braintrust @strands-agents/sdk openai
        ```

        ```bash npm theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
        npm install braintrust @strands-agents/sdk openai
        ```
      </CodeGroup>
    </Step>

    <Step title="Set environment variables">
      ```bash title=".env" theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
      BRAINTRUST_API_KEY=<your-braintrust-api-key>
      OPENAI_API_KEY=<your-openai-api-key>
      ```
    </Step>
  </Steps>

  <h2 id="auto-instrumentation-typescript">
    Auto-instrumentation
  </h2>

  To trace `Agent`, `Graph`, and `Swarm` invocations from `@strands-agents/sdk` without changing your agent construction code, run your app with Braintrust's import hook.

  <Steps>
    <Step title="Initialize Braintrust and invoke your agent">
      <CodeGroup>
        ```typescript title="trace-strands-auto.ts" theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
        import { Agent, tool } from "@strands-agents/sdk";
        import { OpenAIModel } from "@strands-agents/sdk/models/openai";
        import { initLogger } from "braintrust";

        initLogger({
          projectName: "strands-example", // Replace with your project name
          apiKey: process.env.BRAINTRUST_API_KEY,
        });

        const lookupWeather = tool({
          name: "lookup_weather",
          description: "Return the current weather for one city.",
          inputSchema: {
            type: "object",
            properties: { city: { type: "string" } },
            required: ["city"],
          },
          callback: ({ city }) => ({ city, forecast: "sunny" }),
        });

        const agent = new Agent({
          id: "weather-agent",
          name: "weather-agent",
          model: new OpenAIModel({
            modelId: "gpt-5-mini",
            clientConfig: { apiKey: process.env.OPENAI_API_KEY },
          }),
          printer: false,
          systemPrompt: "Answer weather questions concisely.",
          tools: [lookupWeather],
        });

        const result = await agent.invoke("What is the weather in San Francisco?");
        console.log(result.toString());
        ```
      </CodeGroup>
    </Step>

    <Step title="Run with the import hook">
      ```bash theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
      node --import braintrust/hook.mjs trace-strands-auto.ts
      ```

      The auto-instrumentation example uses plain JavaScript so `node --import` can run the file directly. The Braintrust APIs work the same in TypeScript projects — compile your TypeScript to JavaScript, then run the compiled file with the import hook.

      <Note>
        If you're using a bundler, see [Trace LLM calls](/instrument/trace-llm-calls#auto-instrumentation) for plugin and loader setup.
      </Note>
    </Step>
  </Steps>

  <h2 id="manual-instrumentation-typescript">
    Manual instrumentation
  </h2>

  To trace Strands manually, wrap the SDK module with `wrapStrandsAgentSDK` to opt into tracing explicitly.

  <CodeGroup>
    ```typescript title="trace-strands-manual.ts" theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
    import * as Strands from "@strands-agents/sdk";
    import { OpenAIModel } from "@strands-agents/sdk/models/openai";
    import { initLogger, wrapStrandsAgentSDK } from "braintrust";

    initLogger({
      projectName: "strands-example", // Replace with your project name
      apiKey: process.env.BRAINTRUST_API_KEY,
    });

    const { Agent, tool } = wrapStrandsAgentSDK(Strands);

    const lookupWeather = tool({
      name: "lookup_weather",
      description: "Return the current weather for one city.",
      inputSchema: {
        type: "object",
        properties: { city: { type: "string" } },
        required: ["city"],
      },
      callback: ({ city }) => ({ city, forecast: "sunny" }),
    });

    const agent = new Agent({
      id: "weather-agent",
      name: "weather-agent",
      model: new OpenAIModel({
        modelId: "gpt-5-mini",
        clientConfig: { apiKey: process.env.OPENAI_API_KEY },
      }),
      printer: false,
      systemPrompt: "Answer weather questions concisely.",
      tools: [lookupWeather],
    });

    const result = await agent.invoke("What is the weather in San Francisco?");
    console.log(result.toString());
    ```
  </CodeGroup>

  <h2 id="what-traced-typescript">
    What Braintrust traces
  </h2>

  Braintrust captures the Strands span tree for agent and multi-agent runs:

  * Agent spans such as `Agent: weather-agent`, with invocation input, final output, agent ID, agent name, model ID, stop reason, token metrics, and duration.
  * Model spans such as `Strands model: gpt-5-mini`, nested under the agent span, with model metadata, stop reason, token metrics, and latency metrics.
  * Tool spans such as `tool: lookup_weather`, with tool input, output, tool call ID, tool name, status, errors, and duration.
  * Multi-agent orchestration spans (`Strands Graph` and `Strands Swarm`), including per-node spans (`node: <node-id>`), handoffs, status, output, token metrics, and duration.
  * Inline document, image, and video content in span inputs stored as Braintrust attachments rather than embedded bytes.
  * Parent-child nesting under any enclosing Braintrust span.

  <h2 id="resources-typescript">
    Resources
  </h2>

  * [Amazon Strands Agents SDK on npm](https://www.npmjs.com/package/@strands-agents/sdk)
  * [Strands Agent SDK documentation](https://strandsagents.com/)
  * [Trace LLM calls](/instrument/trace-llm-calls)
</View>

<View title="Python" icon="https://img.logo.dev/python.org?token=pk_BdcHD9e5SCW3j1rnJkNyMQ">
  <h2 id="setup-python">
    Setup
  </h2>

  Install the Braintrust SDK and Strands, then set your API keys. The examples below use OpenAI.

  <Steps>
    <Step title="Install packages">
      <CodeGroup>
        ```bash uv theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
        uv add braintrust "strands-agents[openai]" strands-agents-tools
        ```

        ```bash pip theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
        pip install braintrust "strands-agents[openai]" strands-agents-tools
        ```
      </CodeGroup>
    </Step>

    <Step title="Set environment variables">
      ```bash title=".env" theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
      BRAINTRUST_API_KEY=<your-braintrust-api-key>
      OPENAI_API_KEY=<your-openai-api-key>
      ```
    </Step>
  </Steps>

  <h2 id="auto-instrumentation-python">
    Auto-instrumentation
  </h2>

  To trace Strands alongside Braintrust's other supported libraries, call `braintrust.auto_instrument()` before creating your agent.

  <CodeGroup>
    ```python Python theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
    import os

    import braintrust
    from strands import Agent, tool
    from strands.models.openai import OpenAIModel

    braintrust.auto_instrument()
    braintrust.init_logger(
        api_key=os.environ["BRAINTRUST_API_KEY"],
        project="strands-example",  # Replace with your project name
    )


    @tool
    def current_weather(city: str) -> str:
        """Return the current weather for a city."""
        return f"It is sunny in {city}."


    model = OpenAIModel(
        model_id="gpt-4o-mini",
        client_args={"api_key": os.environ["OPENAI_API_KEY"]},
    )

    agent = Agent(
        system_prompt="You answer weather questions concisely.",
        tools=[current_weather],
        model=model,
    )

    result = agent("What is the weather in San Francisco?")
    print(result)
    ```
  </CodeGroup>

  <Accordion title="Trace only Strands">
    To trace Strands without auto-instrumenting other libraries, use `setup_strands()` instead of `braintrust.auto_instrument()`.

    <CodeGroup>
      ```python Python theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
      import os

      from braintrust.integrations.strands import setup_strands
      from strands import Agent, tool
      from strands.models.openai import OpenAIModel

      setup_strands(project_name="strands-example")


      @tool
      def current_weather(city: str) -> str:
          """Return the current weather for a city."""
          return f"It is sunny in {city}."


      model = OpenAIModel(
          model_id="gpt-4o-mini",
          client_args={"api_key": os.environ["OPENAI_API_KEY"]},
      )
      agent = Agent(
          system_prompt="You answer weather questions concisely.",
          tools=[current_weather],
          model=model,
      )

      print(agent("What is the weather in San Francisco?"))
      ```
    </CodeGroup>
  </Accordion>

  <Tip>
    Already running Strands with `StrandsTelemetry` or an OTLP exporter? Keep that setup and route spans to Braintrust using the [Braintrust OpenTelemetry guide](/integrations/sdk-integrations/opentelemetry#python-sdk-configuration).
  </Tip>

  <h2 id="what-traced-python">
    What Braintrust traces
  </h2>

  Braintrust mirrors Strands' native span tree, so the trace shape matches the agent's actual execution. Captured spans:

  * Agent invocation spans (`<agent>.invoke`), with input messages, agent name, model ID, and tool definitions; output stop reason, final message, and structured output.
  * Event loop cycle spans (`event_loop.cycle`), with input messages and cycle ID; output message and tool result message.
  * Model call spans (`<model>.chat`), with input messages, system prompt, and model ID; output message, stop reason, and per-call token usage.
  * Tool call spans (`<tool>.execute`), with tool definition as input; output, execution status, and errors.
  * Parent-child nesting that mirrors Strands' span hierarchy and nests under any enclosing Braintrust span.

  <h2 id="resources-python">
    Resources
  </h2>

  * [Strands Agents SDK documentation](https://strandsagents.com/)
  * [Braintrust Python SDK reference](/sdks/python/versions/latest)
  * [Braintrust OpenTelemetry guide](/integrations/sdk-integrations/opentelemetry)
</View>