Sector ETF Selection with Specialist Agents and Risk Review

# agents example citadel sector pods


Citadel Sector Pods AI Trading Team
===================================

.. meta::
   :description: An AI hedge fund team inspired by Citadel's sector pods: five sector agents pitch ideas, a risk manager pushes back, and a portfolio manager trades. Free Python code.

.. image:: ../docs/assets/ai-trading-team-workflows/citadel-sector-pods.png
   :alt: Five sector agents feed a risk manager and a portfolio manager, then the trade order
   :width: 100%

This AI team is inspired by the "pod" setup used by big multi-manager hedge funds like Ken Griffin's Citadel. Instead of asking one AI to understand the whole market, five specialist agents each cover one part of it and pitch their best sector ETF. A risk manager pushes back, and only the portfolio manager places trades. It shows how specialist agents can check each other before money moves.

How it works
------------

1. **Five sector agents** each study one area: technology and communications, financials, healthcare, energy, and consumer. Each pitches its strongest idea from the ETF universe.
2. **Risk manager agent** reads all five pitches and challenges them: crowded trades, big drops, macro risk, and sudden reversals.
3. **Portfolio manager agent** builds a portfolio across at least three sectors and is the only agent allowed to place orders. The team repeats this once a day.

The copyable example now defaults to GPT-6 Luna with medium reasoning. A leveraged version holds leveraged sector ETFs instead. Earlier BotSpot paper observations used Gemini; changing the model does not change those historical results.

Run it on BotSpot
-----------------

Run this team on `BotSpot <https://botspot.trade/marketplace?utm_source=documentation&utm_medium=docs&utm_campaign=lumibot_ai_examples&utm_content=agents_example_citadel_sector_pods>`_ without installing anything. BotSpot runs LumiBot in the cloud, backtests it, and connects it to your broker.

Backtest tear sheet
-------------------

GPT-6 Luna, January 5 to 16, 2026, Yahoo daily prices, $100,000 start. The team rotated across sector ETFs such as XLC, XLE, XLF, XLI, XLV, and XLB and ended at $101,680 (+2%) while SPY rose 1%. Cash stayed near zero, lowest $33 in the daily stats. Earlier BotSpot paper observations used Gemini and are separate from these Luna backtest results and the current Luna example.

.. image:: ../docs/assets/ai-bot-backtests/citadel-sector-pods.png
   :alt: Backtest tear sheet for the Citadel Sector Pods AI Trading Team
   :width: 100%
   :target: tearsheets/citadel-sector-pods.html

`Open the full tear sheet <tearsheets/citadel-sector-pods.html>`__. A short backtest shows the team works as written. It is not a promise of future returns.

The code
--------

Regular sector ETFs:

.. literalinclude:: ../lumibot/example_strategies/ai_trading_team_citadel_sector_pods.py
   :language: python

Leveraged sector ETFs:

.. literalinclude:: ../lumibot/example_strategies/ai_trading_team_citadel_sector_pods_leveraged.py
   :language: python

Run it yourself
---------------

.. code-block:: bash

   pip install lumibot
   python -m lumibot.example_strategies.ai_trading_team_citadel_sector_pods

Add ``OPENAI_API_KEY`` for GPT-6 Luna and your broker keys (for example ``ALPACA_API_KEY``, ``ALPACA_API_SECRET``, and ``ALPACA_IS_PAPER=true``) to your ``.env`` file. This file trades by default. Set ``IS_BACKTESTING=true`` in your environment to backtest it instead. The default is ``openai/gpt-6-luna``. Set ``AI_TRADING_TEAM_MODEL`` only when deliberately choosing a different model, with its matching provider key.

Good to know
------------

* Inspired by public descriptions of multi-manager hedge funds. Not affiliated with or endorsed by Citadel or its people, and not a copy of any real strategy.
* The picture shows the five sector agents side by side because none of them depends on another. The code calls them one after another.

See :doc:`agents_examples` for more AI trading bots and :doc:`strategy_run_modes` for backtest and live runs.