Introduction

AVM (Agents Virtual Machine) is an advanced execution environment designed to enhance the functionality of Large Language Models (LLMs) and Reasoning Models by enabling secure and efficient code execution.

​

Overview

While LLMs possess extensive knowledge derived from vast code repositories, executing complex computational tasks requires more than natural language processing. This is particularly relevant in sophisticated domains such as data science, research, and financial trading.

AVM provides a secure, isolated environment that integrates seamlessly with existing LLM and AI Agent architectures, granting immediate access to Python-based computational resources.

Python was chosen as the primary execution language due to its extensive representation in LLM training data, allowing for future enhancements, including autonomous model improvement through computational resources.

AVM serves as an AI-optimized alternative to AWS Lambda, offering a streamlined serverless architecture that eliminates infrastructure management complexities.

The operational mechanism is straightforward: Code Input → Computed Output.

Despite its simplified execution model, AVM unlocks powerful capabilities, enabling autonomous agents to perform critical operations such as:

• Transaction signing and cryptographic verification
• API integrations with financial data sources like Coingecko
• Advanced statistical and machine learning computations using industry-standard libraries (e.g., NumPy, scikit-learn)
• Secure execution of computational workloads within isolated environments

​

First Steps

To get started, you can use the API Keys page to generate a new API key. This key will be used to authenticate all requests made to the Sandbox RunsAPI.