Xiamen's first intelligent agent testbed was launched, with the first batch of joint scenario testing projects signed

On the afternoon of October 22, the "Digital-Intelligence Symbiosis Action — Trusted AI Ecosystem Exchange Meeting and the First Batch of Agent Identity Code Issuance Ceremony" was held in Xiamen. At the event, Xiamen's first agent testbed construction project was officially launched.

The testbed is built on the "Model-Data Resonance and Multi-Agent Collaboration Pilot Testing Service Platform." The platform is led by Gaosong Digital (Xiamen) Intelligent Technology Co., Ltd., has been included in the 2026 Xiamen Manufacturing Industry Pilot Testing Service Platform cultivation list, and is Xiamen's first manufacturing pilot testing platform focused on the fields of "model-data resonance" and "heterogeneous agent collaboration." The platform is jointly built with units including the Xiamen Institute of Product Quality Supervision and Inspection, Xiamen University of Technology, Yangmai (Xiamen) Embodied Intelligence Technology Co., Ltd., and the Xiamen Cloud Big Data and IoT Research Institute, providing agents with real-environment verification capabilities such as access, identity authentication, collaborative scheduling, and extreme-condition stress testing.

The testbed's foundational advantage lies in its integrated "software + hardware" pilot testing and verification capability. Gaosong Digital has long participated in national standard verification and platform construction, possessing dual support capabilities in standards and platforms. Joint unit the Xiamen Institute of Product Quality Supervision and Inspection possesses a national-level quality inspection center and an intelligent connected vehicle-road collaboration test site, with testing capabilities for embodied intelligence hardware terminals such as unmanned vehicles and drones, and test scenarios covering simulated scenarios under multiple categories of intelligent connected vehicle regulations and standards. The combination of the software platform and hardware testing resources gives the agent testbed a real "physical examination hall."

The agent testbed launched this time, for the first time, brings together six units to form a complete verification chain: Gaosong Digital provides the AIOS data intelligent operating system as the foundation; Zhongtian Yunzhi provides the model-data space application platform as the entry point; Sanwu Digital provides token metering and operations, responsible for metering; Haichang Shengyao provides heterogeneous large-model Token scheduling, responsible for scheduling; Zongheng Group provides a financial judicial agent, and Hash Technology provides the Carbon Strategy Cloud enterprise dual-carbon agent, as scenarios. From foundation to scenarios, the six units each cover one segment, forming a complete "Data · Model · Agent · Scenario" (D·M·A·S) collaborative verification chain. If one segment is missing, the chain does not connect.

At the event site, joint testing projects for agent application scenarios were also signed. Representatives of the six units took the stage and signed cooperation agreements on projects including the AIOS data intelligent operating system, the financial judicial agent, the green and low-carbon factor library, the agent application store, the AI aggregation platform, and the enterprise agent engine. These projects cover key links such as foundation, entry point, metering, scheduling, and scenarios, meaning that the testbed's verification environment will be fully connected with real industry scenarios, allowing agents to undergo testing in real-world practice.

The testbed and scenario signings fill in a key link of the trusted AI ecosystem. Identity codes address "who it is," agent verification addresses "whether to trust it," the testbed addresses "whether it can run stably," and joint scenario testing addresses "whether it can truly be implemented." The four connect to form a closed loop of "identity—verification—testing—scenario—admission." For agent developers, the testbed is a "training ground" before products enter the industry; for industry users, joint testing is a "real-world inspection" before selection and procurement; for platform operators, testbed and scenario data are the basis for ecosystem admission and risk governance.

The testbed will advance in three stages: "pilot verification—replication and promotion—large-scale operation": first, carry out "small-scope, deep-application" pilots at benchmark enterprises to accumulate the first batch of industry datasets and scenario methodologies; then modularly package industry datasets, industry models, and agent capabilities for batch replication to same-industry and cross-industry scenarios; finally, relying on the "Model-Data Resonance Space" for unified standards and interconnection, achieve normalized large-scale operation, allowing agents to complete access, identity authentication, collaborative scheduling, and extreme-condition stress testing in real environments. Future competition will no longer be just about "who has a larger model," but about who can build the systemic capability for continuous collaborative evolution of data, models, agents, and industry scenarios. The launch of Xiamen's first agent testbed is precisely a key step in this direction.