> For the complete documentation index, see [llms.txt](https://docs.altura.com/altura-documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.altura.com/altura-documentation/altura-intelligence/play-wallet.md).

# Play Wallet

## Overview

\
The Play Wallet is a universal Web3 gaming wallet that integrates seamlessly into all major gaming platforms. It leverages LLMs for user assistance and enforcement of user-defined transaction policies, and deploys graph-based ML models for robust anomaly and fraud detection.

## Model Type & Architecture

* LLM-Based Policy Interpreter: The Play Wallet uses an LLM to interpret human-readable policies (e.g., “Never spend more than 0.1 ETH without confirmation”) and transform them into smart transaction filters.
* Graph-Based ML for Fraud Detection: Graph neural networks (GNNs) and graph-based ML models analyze the transaction graph of cryptocurrency and NFT transfers. Suspicious patterns, unusual activity spikes, and known malicious addresses trigger warnings or require additional verification.
* On-Device Key Management & HSM Integration: The wallet employs secure enclaves or Hardware Security Modules (HSMs) for private key management. The LLM never has direct access to keys; it only guides user interactions and performs risk assessments based on external data.

## Underlying Tech

* Language Modeling + Policy Enforcement: LLMs parse user instructions and convert them into executable filters that run before any on-chain transaction.
* Graph-Based Analytics: GNNs trained on large crypto transaction datasets identify suspicious activity. These models incorporate node embeddings, edge relationships, and historical transaction patterns to flag potential fraud.
* API Integrations: Seamless integration with major blockchains, NFT standards, and asset marketplaces for frictionless asset management.