Kite AI is a decentralised protocol that integrates artificial intelligence into Web3 by enabling users to build, deploy, train, and use AI agents across a permissionless infrastructure. It combines decentralised compute networks, AI model orchestration, and programmable agents into a single composable environment.

The system is structured into three interrelated layers:

• Infrastructure Layer: Provides decentralised compute using GPU providers connected through a resource orchestration layer. This allows for the training and inference of large language models and other machine learning tasks in a trustless and efficient manner. The system supports auto-scaling and allocation based on demand.

• Agent Layer: Supports the creation of on-chain and off-chain agents—AI processes that can interact with users, smart contracts, APIs, and external data sources. Agents can be fine-tuned on user-specific or application-specific datasets and integrated into decentralised workflows.

• Application Layer: Hosts the front-end experiences and protocols that allow users to interact with deployed agents. This includes tools for DAOs, social platforms, data curation, governance assistants, automated trading, and knowledge query interfaces.

Kite positions itself not just as a service layer for AI inference, but as a full-stack environment where agents can be composed, trained, and monetised. It supports multiple runtimes and is chain-agnostic by design, allowing for deployment in Ethereum-compatible environments and other Web3 systems.

The protocol supports modularity in each layer—compute nodes can be run independently, agents can be trained on local data, and application front ends can be built permissionlessly.

Kite (KITE) is the native utility and coordination token of the Kite AI ecosystem. It plays an essential role in securing the network, metering resource use, and incentivising participation across compute providers, model developers, and application builders.

KITE is used in the following ways:

• Compute payment: Users spend KITE to access decentralised GPU infrastructure. This includes training, inference, and multi-agent orchestration.
• Agent hosting: Developers deploy agents using KITE, which covers storage, compute scheduling, and ongoing operations.
• Incentive distribution: Compute node operators, model curators, and contributors receive KITE as compensation for delivering measurable utility to the protocol.
• Governance participation: KITE holders can vote on network parameters, funding allocations, and protocol upgrades.
• Access rights: Some premium tools and orchestration environments require staking or usage of KITE for entry.

The token is designed to support a circular economy in which each participant pays into and receives from the system based on their role. KITE is not purely a payment token—it also acts as a coordination and governance instrument across decentralised AI infrastructure.

• Accessing GPU resources for model training and execution
• Deploying and maintaining AI agents
• Paying for API access and runtime compute
• Staking for priority access or service tiers
• Governance over system upgrades and parameters
• Rewarding data contributors, node operators, and protocol developers

KITE supports metered billing, time-based access, and fine-grained permissions. It can also be used as collateral or access token for advanced orchestration features.