Baidu Intelligent Cloud has launched DuClaw, a zero‑deployment service built to run on the emerging OpenClaw agent ecosystem. The offering arrives preloaded with Baidu’s search, Baidu Baike (its encyclopedia) and academic search “skills,” and is advertised as immediately compatible with a slate of mainstream large models including DeepSeek, Kimi‑K2.5, GLM‑5 and MiniMax‑M2.
The technical pitch is simple: remove integration friction. By deep‑preconfiguring knowledge connectors and retrieval skills, DuClaw promises firms and developers an out‑of‑the‑box pipeline that binds Baidu’s proprietary knowledge sources to external and domestic large language models. The “zero‑deployment” label indicates that customers need minimal engineering effort to expose their agents to Baidu’s indexed content and structured search signals.
For Baidu the move is strategic. It leverages the company’s long‑standing strength in search and encyclopaedic indexing to supply contextual grounding — a remedy for hallucination and factual drift that many LLM deployments still face. Packaging those assets as cloud‑hosted, plug‑and‑play skills also creates a distribution channel for Baidu’s cloud business at a moment when Chinese cloud providers are racing to monetise AI infrastructure beyond raw compute.
DuClaw’s arrival should be read in the context of an accelerating “agentisation” race across China’s tech sector. Competitors from Tencent to ByteDance and a host of startups are building agent runtimes, verticalised assistants and lightweight client experiences; the web page hosting this announcement links to stories of multiple firms launching similar initiatives and of a broader consumer enthusiasm for device‑style agent deployments. That momentum, however, has stirred official scrutiny: Chinese cyber security bodies have issued risk notices related to OpenClaw‑style applications, underscoring concerns about safety, privacy and uncontrolled automation.
The practical implications are twofold. On the one hand, standardised, zero‑setup connectors lower the barrier for enterprises to adopt agents that can access up‑to‑date web knowledge and proprietary internal data. On the other hand, deep integration between a major content owner and third‑party models concentrates a new kind of gatekeeping power: whoever controls the gateway between models and curated knowledge can shape answers, ranking and provenance.
Observers should watch three variables closely: the pace at which large domestic models adopt DuClaw connectors; whether Baidu monetises access to its knowledge skills or bundles them into exclusive enterprise contracts; and how regulators respond to reports of misuse or security lapses. Together these will determine whether DuClaw becomes an interoperable plumbing layer for China’s LLM ecosystem or a battleground for platform dominance and regulatory control.