> ## Documentation Index
> Fetch the complete documentation index at: https://stagehand.readme-i18n.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent

> Web AI agents for any task

<CodeGroup>
  ```typescript TypeScript theme={null}
  const agent = stagehand.agent();
  await agent.execute("Sign me up for a library card");
  ```

  ```python Python theme={null}
  # Native Stagehand agent support for Python is coming soon.
  # agent = stagehand.agent()
  # await agent.execute("Sign me up for a library card")
  ```
</CodeGroup>

Software has always been deterministic and repeatable, but with AI agents it's difficult to replicate a workflow. **Stagehand combines the best of both worlds: intelligence and determinism**.

Web Agents in Stagehand are fully customizable. You can use any LLM / VLM / Computer Use provider, set system prompts, add custom tools, and more.

<CodeGroup>
  ```typescript TypeScript theme={null}
  // For Computer Use Agent (CUA) models
  const agent = stagehand.agent({
    provider: "openai",
    model: "computer-use-preview",
    instructions: "You are a helpful assistant that can use a web browser.",
    options: {
      apiKey: process.env.OPENAI_API_KEY,
    },
  });

  // You can define additional options like max_steps and auto_screenshot
  await agent.execute({
    instruction="Sign me up for a library card",
    max_steps=20,
  })
  ```

  ```python Python theme={null}
  # For Computer Use Agent (CUA) models
  agent = stagehand.agent({
      model="computer-use-preview",
      instructions="You are a helpful assistant that can use a web browser.",
      options={
        "apiKey": os.getenv("OPENAI_API_KEY"),
      },
  })

  # You can define additional options like max_steps and auto_screenshot
  result = await agent.execute({
      instruction="Sign me up for a library card",
      max_steps=20
  })

  ```
</CodeGroup>

### **Arguments:** [`AgentOptions`](https://github.com/browserbase/stagehand/blob/main/types/agent.ts)

<Tabs>
  <Tab title="TypeScript">
    <ParamField path="provider" type="string" required>
      Specifies the LLM provider.
      Currently supporting:

      * **OpenAI**:  `openai`
      * **Anthropic**:  `anthropic`
    </ParamField>

    <ParamField path="model" type="string" required>
      Specifies the LLM model.
      Currently supporting:

      * **OpenAI**:  `computer-use-preview`
      * **Anthropic**:  `claude-sonnet-4-20250514` and `claude-3-7-sonnet-latest`
    </ParamField>

    <ParamField path="instructions" type="string">
      Specifies the system prompt
    </ParamField>

    <ParamField path="options" type="object">
      Configuration options for the LLM provider. Currently supports OpenAI and Anthropic SDK client options.
    </ParamField>
  </Tab>

  <Tab title="Python">
    <ParamField path="model" type="string" required>
      Specifies the LLM model.
      Currently supporting:

      * **OpenAI**:  `computer-use-preview`
      * **Anthropic**:  `claude-sonnet-4-20250514` and `claude-3-7-sonnet-latest`
    </ParamField>

    <ParamField path="provider" type="string" optional>
      Specifies the LLM provider.
      Currently supporting:

      * **OpenAI**:  `openai`
      * **Anthropic**:  `anthropic`
    </ParamField>

    <ParamField path="instructions" type="string">
      Specifies the system prompt
    </ParamField>

    <ParamField path="options" type="dict">
      Configuration options for the LLM provider. Currently supports OpenAI and Anthropic SDK client options.
    </ParamField>
  </Tab>
</Tabs>

### **Returns:** [`Promise<AgentResult>`](https://github.com/browserbase/stagehand/blob/main/types/agent.ts)

<ParamField path="success" type="boolean">
  If the agent execution completed without errors
</ParamField>

<ParamField path="message" type="string">
  Summary or error message from the agent
</ParamField>

<ParamField path="actions" type="AgentAction[]">
  List of actions performed during execution
</ParamField>

<ParamField path="completed" type="boolean">
  Whether the agent fully completed its task (can be false even if success is true)
</ParamField>

<ParamField path="metadata" type="Record<string, unknown> | dict">
  Optional additional data about the execution
</ParamField>