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

# Quick Start

> Build and run your first orxhestra agent in under five minutes — as a YAML file you run from the terminal, or as Python code.

Pick the path that matches how you want to work. The YAML tab is the shortest route for most users — the SDK tab is for when you need full programmatic control.

<Tabs>
  <Tab title="YAML (Composer)">
    <Steps>
      <Step title="Install">
        ```bash theme={null}
        pip install 'orxhestra[cli,anthropic]'
        export ANTHROPIC_API_KEY=sk-ant-...
        ```
      </Step>

      <Step title="Save orx.yaml">
        ```yaml title="orx.yaml" theme={null}
        defaults:
          model:
            provider: anthropic
            name: claude-opus-4-7

        agents:
          assistant:
            type: llm
            instructions: "You are a helpful weather assistant."

        main_agent: assistant
        ```
      </Step>

      <Step title="Run it">
        ```bash theme={null}
        orx orx.yaml
        ```

        Drops you into an interactive REPL with your agent as the root. Type a question, watch the stream. `/help` lists the slash commands; `/exit` when done.
      </Step>
    </Steps>

    <Tip>
      Want to expose the same agent as an HTTP server?

      ```bash theme={null}
      orx orx.yaml --serve -p 9000
      ```

      Serves A2A v1.0 JSON-RPC on `:9000`. See [Integrations → A2A](/integrations/a2a) for the wire format.
    </Tip>

    <Tip>
      Want signed events and an audit log? Opt in with a single flag or YAML block:

      ```bash theme={null}
      orx identity init --path ./keys/agent.key
      orx orx.yaml --identity ./keys/agent.key
      ```

      The banner and `/session` command show the active signer DID. See [Composer → Identity, trust, and attestation](/composer/overview#identity-trust-and-attestation).
    </Tip>
  </Tab>

  <Tab title="Python SDK">
    ```bash theme={null}
    pip install 'orxhestra[anthropic]'
    export ANTHROPIC_API_KEY=sk-ant-...
    ```

    ```python title="basic_agent.py" theme={null}
    import asyncio
    from langchain_anthropic import ChatAnthropic
    from langchain_core.tools import tool

    from orxhestra import LlmAgent, Runner, InMemorySessionService


    @tool
    def get_weather(city: str) -> str:
        """Get the current weather for a city."""
        return f"The weather in {city} is sunny and 22 degrees."


    agent = LlmAgent(
        name="WeatherAgent",
        model=ChatAnthropic(model="claude-opus-4-7"),
        tools=[get_weather],
        instructions="You are a helpful weather assistant.",
    )

    runner = Runner(
        agent=agent,
        app_name="demo",
        session_service=InMemorySessionService(),
    )


    async def main():
        async for event in runner.astream(
            user_id="user-1",
            session_id="session-1",
            new_message="What's the weather in Copenhagen and Berlin?",
        ):
            if event.is_final_response():
                print(event.text)


    asyncio.run(main())
    ```

    ```bash theme={null}
    uv run python basic_agent.py
    ```
  </Tab>
</Tabs>

## Next steps

<Cards>
  <Card title="Compose a team" icon="wand-magic-sparkles" href="/composer/overview">
    Multi-agent YAML with loops, parallel fan-out, and transfer routing.
  </Card>

  <Card title="Agent types" icon="sitemap" href="/concepts/agents">
    LlmAgent, ReActAgent, and the four composite wrappers.
  </Card>

  <Card title="Tools" icon="wrench" href="/tools/overview">
    Function tools, filesystem, shell, MCP, and custom tool types.
  </Card>

  <Card title="Trust & audit" icon="shield-check" href="/composer/overview#identity-trust-and-attestation">
    Sign events, verify peers, persist audit logs — all opt-in.
  </Card>
</Cards>