- Lingguang transitions from passive AI assistant to proactive personal agent with long-term memory and autonomous multi-step task execution.
- The new version uses Skill and MCP to enable cross-app operations including DingTalk messaging and smart home device control.
- Beta use cases target domestic pain points such as market movement tracking, school check-in management, and real-time travel itinerary replanning.
- Ant Group explicitly benchmarks Lingguang against Meta Muse, signaling intensified competition in the global AI agent race.
From Chat Assistant to Proactive Agent
Ant Group has begun small-scale internal beta testing of a significantly upgraded version of Lingguang, its consumer-facing AI product. According to project insiders cited by AIBase, the new iteration marks a fundamental shift from a basic conversational assistant and application generator to what the company describes as an AI-native personal agent capable of maintaining long-term memory and proactively executing complex, multi-step tasks on behalf of users.
Preparation for the agent-oriented redesign began approximately six months ago, with the team iterating through multiple versions ahead of the current beta. A full public release is planned for October 2026.
Cross-App Execution via Skill and MCP
The upgraded Lingguang leverages two integration mechanisms, Skill and MCP, to interact with external applications and services. During the beta, demonstrated capabilities include sending messages through Alibaba's enterprise collaboration platform DingTalk and controlling connected smart home devices, indicating a push toward full-workflow task delegation rather than single-turn query responses.
These cross-app operations represent a move toward an architecture where the agent can coordinate actions across multiple domestic platforms.
Domestic Use Cases Drive Beta Design
The beta program focuses on scenarios involving long-duration, multi-variable challenges that traditional AI assistants struggle to handle. Cited examples include helping office workers track financial market movements throughout the trading day, assisting parents in managing their children's daily DingTalk school check-ins, and automatically reorganizing travel itineraries when flights are delayed or scenic areas reach capacity limits.
Benchmarking Against Meta Muse
Ant Group has positioned the new Lingguang directly against Meta Muse, the personal AI agent Meta launched in September 2026. This explicit benchmarking reflects both the competitive pressure from U.S. tech giants in the agent space and Ant's confidence that its product can match Western counterparts while serving distinctly Chinese user workflows and platform ecosystems.
The launch adds Ant Group to a growing roster of major Chinese technology companies accelerating investment in proactive AI agents and full-workflow task delegation.
Sources & context
Go to the original material. Company claims remain attributed to their sources.
01
