The construction of the country’s first trusted AI agent testbed was launched in Nanjing

Recently, at the 14th plenary meeting of the Chinese Artificial Intelligence Industry Development Alliance (AIIA) and the Artificial Intelligence Empowerment New Industrialization In-depth Tour (Nanjing Station), the construction of the country’s first trusted AI agent testbed was officially launched.

In recent years, agents, as the key application mode of large models, can transform model capabilities into practical problem-solving capabilities, showing a vigorous development trend. The China Academy of Information and Communications Technology (CAICT) attaches great importance to the development of agent technology and application, promotes the high-quality development of agents around technical research, standard development, evaluation and testing, etc., and recently launched the solicitation of joint construction units of agent testbeds, which has received great attention and support from enterprises and experts in the industry.
This time, the China Academy of Information and Communications Technology, together with China Unicom Digital Technology, China Mobile Research Institute, Huawei, Baidu, VIVO, ZTE and other units, launched the construction of a trusted AI agent testbed in Nanjing, aiming to create an integrated test environment and comprehensively verify and optimize the agent’s ability to understand and analyze, plan and make decisions, communicate and interact, and perform tasks. In terms of technical architecture, at the resource allocation layer, the system configures the basic resources of the agent testbed, including computing facilities, data resources, network facilities, operating systems, basic software, etc. At the environment construction layer, the agent test environment is mainly built, including basic test environments such as running script debugging and data visual analysis, as well as operation and management environments such as model access and agent deployment. In the task test layer, the test system is built for typical tasks such as MCP protocol verification test, single-agent benchmark test, and multi-agent collaborative test.

In the next step, Nanjing will rely on the opportunity of the construction to promote the introduction of the main resources for the construction of trusted AI agent test beds, and attract more AI agent enterprises to deploy and develop in Nanjing. At the same time, we will promote Nanjing artificial intelligence enterprises to actively support and participate in the construction of test beds, and contribute more Nanjing strength to the development of the national artificial intelligence industry.