Recently, the Cangjie community officially released the Cangjie Magic Agent development framework. This is reportedly the first Large Language Model (LLM) Agent development platform natively built upon Huawei's self-developed Cangjie programming language.
The Cangjie Magic framework aims to reconstruct the paradigm of agent development through three core technological breakthroughs: firstly, a unique Agent DSL (Domain-Specific Language) architecture, designed based on the characteristics of the Cangjie language, enabling declarative programming for agent modeling; secondly, native support for the MCP (Model Context Protocol) communication protocol; and thirdly, an intelligent scheduling engine integrating a modular service call mechanism and a dynamic task planning algorithm.
Its core component, Cangjie Agent DSL, is a language specifically designed for defining and managing Agents. Developers can enhance the functionality and flexibility of Agents through structured system prompts, tools, and various collaboration strategies, achieving full lifecycle management from agent definition and behavior orchestration to runtime monitoring. Cangjie Agent DSL is designed as an embedded DSL (eDSL) for the Cangjie language, ultimately being translated into regular Cangjie code and compiled by the Cangjie compiler.
Currently, the Cangjie Magic framework has achieved full platform adaptation for HarmonyOS, Windows, macOS, and Linux systems, initially forming a closed-loop desktop ecosystem. More noteworthy is the official roadmap, which indicates plans to implement native agent call capabilities for Android and iOS in the third quarter of this year, leveraging the cross-platform compilation capabilities of the Cangjie language. This will significantly expand the application scenarios of agents on mobile devices.
To help developers get started quickly, the Cangjie AI Agent framework team has also written a detailed user manual, using feature introductions and example demonstrations to help developers master the essence of Cangjie Agent DSL and apply it to practical development.