CARV launches D.A.T.A framework: Building cross-chain data infrastructure for AI Agents

Author: Haotian

Recently, CARV released a D.A.T.A framework and standard. As the name suggests, Virtual's G.A.M.E is a development and deployment framework focused on gaming scenarios, while D.A.T.A is a data framework面向通用 "chain-like" scenarios, primarily addressing enhancements in cross-blockchain data processing, privacy computing, and automated decision-making capabilities for AI Agent data interactions.

Below is an analysis of D.A.T.A in comparison to the G.A.M.E framework:

1) The G.A.M.E framework provided by Virtual helps developers create AI Agents capable of autonomous planning and decision-making within game environments. Its primary target is LLMs (large language models).

It enables large models to achieve autonomous decision-making and action planning based on natural language input through a fine-tuned high-level planner (HLP) and low-level planner (LLP). The HLP formulates strategies and tasks, while the LLP translates these tasks into concrete executable actions. Ultimately, it allows developers to quickly build and deploy production-ready AI Agents using modular components—for example, providing intelligent decision-making for NPCs or players in games.

In contrast, the D.A.T.A framework provided by CARV is a "data" infrastructure designed for general-purpose scenarios, aiming to provide high-quality on-chain and off-chain data support for AI Agents. Its main focus is enhancing AI Agents' inter-chain "data" communication and interaction capabilities.

As a modular, highly extensible general-purpose blockchain, its SVM Chain introduces a cross-chain data standardization protocol that enables AI Agents to uniformly access and process data from different blockchains. Meanwhile, the blockchain's verifiable and traceable mechanisms ensure data security during transmission and processing. Additionally, the integration of TEE and ZK technologies guarantees data privacy. Clearly, CARV primarily defines a mechanism enabling AI Agents to operate interoperably across chains.

2) How exactly does it work? The CARV ecosystem, tailored for AI Agent inter-chain interactions, consists of four core components: SVM Chain, D.A.T.A Framework, CARV_ID, and CARV_Labs;

Refer to documentation for details

1. SVM Chain provides the underlying blockchain infrastructure, including handling cross-chain transactions, supporting smart contract execution, maintaining consensus mechanisms, and other foundational functions—essential chain infrastructure required for the D.A.T.A framework to operate;

2. The D.A.T.A framework and standards, which mainly include cross-chain data standardization, data aggregation and parsing, and privacy computing support. This process involves retrieving raw data from SVM Chain, associating it via the ID system and Agent identity system, and ultimately delivering standardized data to the application layer;

3. CARV_ID identity management system, implemented based on the ERC7231 standard, covering AI Agent identity tagging, authentication, permission management, data authorization, etc., working collaboratively with the D.A.T.A framework for data management;

4. CARV_Labs, primarily providing foundational support through project incubation, ecosystem application deployment, and technological innovation incentives, enabling AI Agent applications supported by other technical framework modules to achieve real-world implementation.

In summary, it is clear that CARV's approach to the AI Agent赛道 leverages the inherent advantages of its chain-like architecture, focusing on the critical functionality of on-chain and off-chain data processing required for AI Agent operations. By aggregating data, defining data standards, and building data verification and traceability mechanisms, CARV aims to become a blockchain architecture fully compatible with AI Agents.

The G.A.M.E and D.A.T.A frameworks are fundamentally different—one deeply explores AI Agents' autonomous decision-making and action execution capabilities within gaming contexts, enabling more efficient translation of natural language inputs into in-game actions, while the other spans multi-chain environments, taking AI Agents' chain-related needs as its guiding principle and using "data" as the entry point, positioning CARV as the first general-purpose infrastructure chain serving AI Agents.