NEAR AI Hub is a general term for the AI-focused direction of the NEAR ecosystem, combining infrastructure for autonomous agents, confidential inference, Agent Market, and blockchain-based settlement mechanisms. The project is developing the concept of user-owned AI, where agents are designed to act in users' interests, handle data securely, and independently perform economic actions when required. In 2026, this strategy gained practical components including IronClaw, NEAR AI Cloud, Agent Market, and integration with NEAR Intents. As a result, NEAR is used not only as a blockchain for smart contracts but also as a settlement and coordination layer for the AI agent economy.
Contents
- NEAR AI Hub and the User-Owned AI Concept
- Confidential AI Agents
- NEAR AI Agent Market and the Autonomous Agent Economy
- NEAR Intents, NEAR Token and Blockchain Infrastructure
- NEAR AI Development and the Future of the Agent Economy
1. NEAR AI Hub and the User-Owned AI Concept
NEAR's connection with artificial intelligence goes back to the origins of the protocol itself. NEAR co-founder Illia Polosukhin previously participated in Google research that contributed to the development of the Transformer architecture. Following the growth of NEAR Protocol, AI once again became a central part of the ecosystem's strategy, with NEAR AI focusing on infrastructure for consumer and enterprise AI agents.
The core concept is called user-owned AI. Unlike centralized AI services, where requests, computation, and system management are controlled by a single provider, NEAR proposes an architecture in which users retain control over their data, assets, and agent actions. In this model, blockchain is primarily used for identity, payments, asset ownership, and economic transactions.
The term NEAR AI Hub does not refer to a separate blockchain, L2 network, or new token. In practice, it describes a set of interconnected NEAR AI products and services. Key components of the current architecture include NEAR AI Cloud for confidential computing, IronClaw for running agents, and Agent Market for interactions between customers and specialized AI systems.
An important part of the strategy is the transition from chatbots to agents. A conventional language model primarily responds to prompts, while an agent can work with external tools, APIs, wallets, and services, execute sequences of actions, and continue tasks without constant manual supervision. This autonomy expands AI capabilities but also increases requirements for credential protection, permissions, and verifiable execution.
2. Confidential AI Agents
One of NEAR AI's core products is IronClaw, an open-source runtime for autonomous AI agents written in Rust. It is designed to perform long-running tasks, interact with tools, and automate workflows. In February 2026, NEAR AI introduced IronClaw alongside an expanded confidential computing infrastructure.
A key distinction of the architecture is its use of Trusted Execution Environments. An agent and its data can operate inside a hardware-isolated environment that restricts access by the infrastructure operator. NEAR AI uses Intel TDX and NVIDIA Confidential Computing technologies, while hardware attestation makes it possible to verify that computations were actually performed within the specified protected environment.
Key components of the NEAR AI infrastructure include:
- IronClaw — an open-source runtime for autonomous AI agents;
- NEAR AI Cloud — infrastructure for private inference and agent hosting;
- Trusted Execution Environments for hardware-level isolation of data and code;
- hardware attestation for verifying execution environments;
- support for open and user-selected AI models;
- Agent Market for discovering and hiring specialized agents;
- NEAR Intents for payments and cross-chain operations;
- NEAR Protocol as a blockchain settlement and coordination layer.
Confidential inference is particularly important for agents working with financial, corporate, or personal data. If an agent has access to email, documents, APIs, or payments, relying solely on a provider's privacy policy may not be sufficient. TEE reduces the need to trust the server operator by protecting data directly during computation.
However, TEE does not automatically make an AI agent secure. Model errors, prompt injection, incorrectly configured permissions, and vulnerabilities in external tools remain separate risks. Confidential computing should therefore be viewed as one layer of protection rather than a replacement for access controls, action auditing, and restrictions on autonomous agent operations.
3. NEAR AI Agent Market and the Autonomous Agent Economy
In February 2026, NEAR AI introduced Agent Market, a marketplace where AI agents can discover paid tasks and compete to perform them. Users define a task and its conditions, after which registered agents can propose a price and timeframe. Payment is locked through an escrow mechanism and transferred to the provider after the work is completed.
Agent Market differs from a conventional AI application directory because agents act as independent economic participants. They can discover tasks, submit proposals, and receive payments. The current marketplace includes agents for research, programming, data analysis, security, tool-based workflows, and other tasks, while the public catalog displays reputation and activity history.
| Component | Primary Function | Role for AI Agents | Blockchain Role |
|---|---|---|---|
| IronClaw | Agent execution and management | Runtime environment | Integration with the NEAR ecosystem |
| NEAR AI Cloud | Confidential inference | Protected computation | Support for agent-native infrastructure |
| Agent Market | Marketplace for tasks and agents | Finding providers and paid work | Escrow and settlements |
| NEAR Intents | Cross-chain execution | Automated transactions | Moving assets between networks |
| NEAR | Native network asset | Economic operations | Staking, fees, and AI mechanisms |
As the marketplace developed, its settlement model evolved. Initially, tasks and payouts were closely connected to NEAR, while later versions of the market added USDC settlements and escrow mechanisms to protect participants. This demonstrates that NEAR's agent economy is not limited to mandatory payments exclusively in the native token.
The economic autonomy of AI agents also creates new challenges. It is necessary to determine who is responsible for incorrect results, how reputation is established, and under what conditions payments can be refunded. The more access an agent receives to external services and financial operations, the more important limited permissions, activity logging, and the ability to revoke access become.
4. NEAR Intents, NEAR Token and Blockchain Infrastructure
NEAR Intents serve as a bridge between AI agents and cryptocurrency infrastructure. In a conventional blockchain transaction, the user specifies the exact sequence of actions. An intent instead describes the desired outcome, while specialized participants known as solvers compete to execute the request as efficiently as possible.
This is particularly relevant for AI agents because they do not need to manually manage every intermediate operation. An agent can define a final objective, such as exchanging an asset or making a payment, while the Intents infrastructure organizes execution. NEAR positions this mechanism as a universal transaction layer for the AI economy and cross-chain interactions.
NEAR Intents also reduces an agent's dependence on a single network. The infrastructure is designed to work with assets across multiple blockchains and abstracts some of the complexity associated with bridges, liquidity, and gas. This aligns with NEAR's broader strategy, in which AI agents can interact with a multichain environment rather than operating exclusively within NEAR Protocol.
The native NEAR token retains its fundamental network functions, including staking, security, and network operations. In July 2026, NEAR AI also launched a staking-based model for accessing confidential inference and agent hosting. Users lock NEAR and receive AI compute credits, while the underlying stake remains owned by the user and can be withdrawn according to the system's conditions.
This model connects an existing crypto asset with AI infrastructure without introducing a separate NEAR AI token. At the same time, staking NEAR does not provide direct ownership of AI models or guarantee revenue from the agent market. Its role depends on the specific product: on the main network it contributes to security, while within NEAR AI it can serve as a mechanism for obtaining compute credits.
5. NEAR AI Development and the Future of the Agent Economy
In 2026, NEAR AI expanded its infrastructure for confidential agents. In addition to IronClaw and Agent Market, it introduced the Confidential GPU Marketplace and multimodal private inference. The infrastructure also gained integrations with external AI products, including Venice.
In July, private inference was integrated into Corbits for enterprise multi-agent workflows, while in August NEAR AI Cloud added Intel Trust Authority for independent attestation verification. This makes it possible to verify a protected execution environment without having to rely entirely on claims made by the infrastructure operator.
NEAR's strategy combines cross-chain infrastructure with autonomous AI agents. NEAR Intents handles operations involving digital assets, while NEAR AI, IronClaw, and confidential computing form the agent layer. Together, these components create a model in which AI can analyze data and perform authorized economic actions.
The main risks are related to security and the actual quality of agent autonomy. AI agents can make mistakes, misinterpret tasks, or encounter malicious external content. TEE protects computation but does not guarantee the correctness of model decisions, while blockchain provides verifiable settlements but does not guarantee the quality of completed work.
NEAR AI Hub is therefore better understood as an evolving AI layer within the NEAR ecosystem rather than a separate blockchain or token. IronClaw, NEAR AI Cloud, Agent Market, and NEAR Intents form infrastructure for an agent economy in which AI systems can perform tasks and interact with digital assets.
