Over the years, we have built a rich suite of tools and APIs to help you build your investment strategy. You can truly build virtually any type of strategy on QuantConnect; however, having so many features can make it challenging to learn quickly, and requiring knowledge of Python and our API increases the barrier to traders and investors using QuantConnect.

Today, we’re officially releasing "Mia V2", an AI quant developer assistant that can independently ideate strategies, write the code for them, and run the backtests to debug them. With Mia V2, you have the full power and flexibility of code, combined with the user-friendliness of a chat-based interface. The agentic peer programmer is hosted in our online algorithm lab; with no downloads or separate accounts needed. Check out the demonstration below:

What can V2 do? 

  • Idea Generation: Suggestions for new strategies or improvements to your current strategy.
  • Quant Developer/Peer Programming: Implementation of your ideas using any of the QuantConnect API tools available.
  • Debugging Assistant: Understanding runtime bugs and fixing code issues directly that cause them - and then repeating the backtest process to solve all bugs.
  • Optimization of Speed and Performance: Review code structure and design considerations to identify opportunities for improving backtesting speed.
  • Review Strategy for Improvements: Improving performance through strategy enhancements. 

How is it different from V1?

Two years ago, we released a “workflow” based agent (Mia V1) who was able to provide answers to people’s questions quickly, but could not execute the changes suggested, or adapt to the various needs of the community.

Now with her fully agentic mode, Mia V2 can adapt and perform any quant development work supported within the QuantConnect platform. 

How does it work?

The new Mia is a tight wrapper over the leading US LLM models, paired with a cloud-only MCP. By creating a cloud-only MCP, we are able to remain compliant with our data licenses and still give the community powerful new tooling.

We worked diligently to simplify the MCP, reducing token usage and enhancing the quality of LLM responses. Our earlier work on the LEAN Python stubs contributed significantly to high-quality syntax checking, and informative warnings. When paired with our MCP, it significantly aided the LLM in writing code.

Getting Started

To get started, write a strategy description in the “Ask Mia” tab on the right of the IDE. Bronze, Silver, and Gold support plans include free premium Mia credits each month, allowing you to use her full power.

If you need additional LLM tokens beyond the support packages, you can purchase them via the QCC Billing page. We pass through the token cost of the major models, so the consumption of use in QCC vs use with the models directly should be similar. By removing previous fixed question limits you can now use Mia as much as you need.

The cost of conversations is recorded on your organization's billing page, where you can see the  QCC consumed. Generally, projects with extensive code and lengthy discussions are more expensive. For more information, refer to the new documentation, 'AI Strategy Development'.

Mia V2 represents a significant leap forward in ease of use for QuantConnect, and we’re excited to see her help bring quantitative finance to the broader investment community.

Happy Coding!

QuantConnect Team