Beijing has introduced a sweeping policy package designed to accelerate the development of intelligent agents, positioning the city as a global hub for artificial intelligence innovation.

The plan, jointly issued by several municipal agencies and approved by the Beijing Municipal Government, focuses on strengthening AI infrastructure, promoting AI-native applications, supporting one-person AI companies, and building a "token economy" centered on AI services and inference efficiency. The policy document is available on the Beijing Municipal Government website.

AI Agents as a New Growth Driver

The initiative aims to advance intelligent agents capable of learning, reasoning, planning, and completing tasks with greater autonomy. Authorities stated that the effort supports China's national AI strategy and seeks to unlock new opportunities within the intelligent economy.

AI agents have become a major focus of AI development because they are designed to move beyond single-turn chatbot responses toward task execution across software tools, enterprise workflows, devices, and services. Beijing's plan reflects that shift by treating agents not only as model capabilities, but also as a layer of applications, infrastructure, standards, and governance.

The plan calls for progress in large AI models through enhancements in online learning, long-context reasoning, multi-agent collaboration, memory systems, and world models. It also encourages the development of shared AI infrastructure, open-source frameworks, interoperability standards, and secure AI technology systems.

AI-Native Software, Smart Devices, and Enterprise Adoption

The measures encourage companies to redesign software products around intelligent agents, including research assistants, enterprise AI operating systems, and AI-powered industrial solutions.

Beijing also intends to integrate intelligent agents into next-generation hardware such as smartphones, smart glasses, robots, vehicles, and wearable devices. Eligible AI-powered products may receive support through consumer trade-in programs and digital product subsidies.

For enterprises, the policy emphasis on AI-native systems points to a broader push to embed models into daily operations rather than treat them as standalone tools. That makes interoperability, data security, reliability testing, and measurable service quality important parts of adoption, especially when agents are used in industrial, office, or consumer-facing scenarios.

Support for One-Person AI Companies

A central component of the policy is backing for One-Person Companies (OPCs), a business model enabled by AI-driven automation.

The city plans to simplify company registration, provide professional incubation services, establish dedicated OPC communities, and build public platforms offering computing resources, financing, intellectual property assistance, and entrepreneurship support. Authorities also intend to connect AI-native startups with larger enterprises through supply-demand matching platforms.

The OPC framework is notable because it links AI automation with entrepreneurship policy. By pairing simplified registration with access to computing power, financing, and intellectual property services, Beijing is attempting to lower the operational barriers for small AI-native businesses that may rely heavily on automated product development, customer service, content generation, or workflow management.

Developing an AI Token Economy

The policy introduces support for a "token economy." In this context, tokens refer to AI computational units rather than blockchain-based cryptocurrencies.

Beijing intends to promote technologies that improve inference efficiency and reduce computing costs. The city also aims to develop new service models, including Token as a Service (TaaS), Agent as a Service (AaaS), and Result as a Service (RaaS).

Additionally, the city will explore standards for token pricing and service quality, and will develop evaluation systems based on AI performance, usage levels, and conversion efficiency.

This focus reflects a practical constraint in deploying large AI systems: inference can become a recurring cost as users and enterprises run models at scale. Standardizing how tokens are priced, measured, and tied to service quality could help providers and customers compare AI services more consistently, while efficiency improvements may expand the range of applications that can be economically deployed.

Expanding Infrastructure, Security, and Open-Source Development

To support the initiative, Beijing will expand computing infrastructure through its "Galaxy Computing Corridor." The city will also provide computing and token vouchers, establish AI security testing platforms, and create regulatory sandboxes.

Beijing plans to strengthen AI talent development and broaden international cooperation. The city will also support open-source communities and encourage developers to contribute AI frameworks, operating systems, and interoperability protocols to the wider ecosystem.

The inclusion of security testing platforms and regulatory sandboxes shows that the city is pairing deployment support with mechanisms for evaluation and oversight. For developers and companies, the areas to watch include how voucher programs are implemented, how token-related standards are defined, and which agent applications are prioritized for enterprise, consumer, and public-service use.