Shenzhen’s Longgang district has moved swiftly to turn a viral open‑source project into an economic strategy. After OpenClaw — an open‑source AI agent framework that surged in popularity during China’s Two Sessions — captured public and political attention, Longgang’s Human‑Machine Bureau published a ten‑point support package, dubbed the “Lobster Ten,” aimed at making the district the world’s preferred home for intelligent‑agent development and one‑person companies (OPCs).
The Human‑Machine Bureau, a government office created in 2025 to coordinate artificial intelligence and robotics policy, has positioned itself as a single entry point for industry needs. By concentrating functions that previously cut across multiple agencies — from industrial planning and scene opening to data and compute provisioning — the bureau claims it can implement fast, flat decision‑making and “close the loop” between policy, industry and real‑world scenarios.
OpenClaw and the associated OPC entrepreneurship model epitomize a shift in China’s AI ecosystem. The government’s recent inclusion of “intelligent agents” in the national work report signaled official recognition that agent‑based software could become a new infrastructure for economic activity. Longgang’s measures explicitly target this transition: subsidies, ecosystem supports and operational services are tailored to help developers build, deploy and commercialize agents without first joining a large platform.
The package is not merely promotional. Longgang articulates quantitative ambitions — to see agent technologies deeply applied in more than 80% of the district’s key industries, to incubate 100 full‑scale intelligent application flagships and to grow the AI sector to over RMB100 billion by 2027 — while building what it calls a national AI open‑source community and OPC‑friendly entrepreneurial environment.
Officials stress that speed is paired with security. Longgang’s plan sets out a three‑tiered safety architecture covering technology, management and ecosystem safeguards: mandatory isolation of government agent deployments from the public internet, strict least‑privilege and human‑in‑the‑loop controls, supply‑chain audits for agent skill‑packages and routine emergency drills. The district is also promoting “security‑as‑a‑service” offerings from cloud providers and pushing for domestic stack components to reduce external dependencies.
The Longgang initiative crystallizes broader tensions and opportunities in China’s AI trajectory. On the one hand, open‑source agents and OPCs promise a decentralised, democratized innovation model that could loosen the near‑monopoly of big tech on advanced AI applications and accelerate real‑world adoption. On the other, a patchwork of aggressive local policies risks fragmenting standards, creating local winners who “lock in” developers within their regulatory and commercial ecosystems.
Practical obstacles remain. Attracting and retaining global talent, ensuring interoperability between agents and existing platforms, navigating intellectual property and export controls, and balancing openness with rigorous security will test Longgang’s blueprint. Whether a district‑level approach can scale without national harmonization — and whether its incentives will encourage long‑term, rather than speculative, enterprise — are open questions.
Internationally, Longgang’s experiment is a bellwether. If the district succeeds in aggregating developers, capital and deployment scenarios around an open‑source agent stack, it will offer a replicable model of local industrial policy for software‑defined economic growth. If it fails, the episode may instead underscore the limits of municipal experimentation in an industry shaped by platform dynamics and geopolitics.