仓颉AI编程助手
仓颉AI编程助手是一款轻量化VS Code智能AI编程插件,内置强大AI智能体能力,深度融入编辑器日常开发流程,支持自然语言写代码、代码自查优化、终端命令运行、浏览器自动化、实时代码行内补全等全能开发功能,适配本地私有化大模型与主流云端AI模型,大幅提升编码效率。
Cangjie Code Assistant
Cangjie Code Assistant is an open-source AI coding agent designed to boost your daily development workflow. It turns natural language into working code, automates repetitive tasks, runs terminal commands, and provides intelligent inline suggestions — all directly inside VS Code.
产品概述
- 智能AI编程:自然语言快速生成代码、智能重构、上下文调试、实时行内代码补全
- 全流程自动化:编辑器内直接执行终端命令、浏览器流程自动化,无需切换工具
- 全模型兼容:支持市面主流大语言模型,自由切换适配个人、团队、私有化部署场景
- 自定义开发模式:内置多种开发工作模式,适配需求编写、代码开发、代码审核、问题调试等各类场景
Overview
- AI-powered development: Generate code from prompts, refactor automatically, debug with context, and get intelligent inline completions.
- Automation & tooling: Run terminal commands, automate browser interactions, and streamline your workflow without leaving the editor.
- Broad model support: Works with over 500 AI models including Claude, Gemini, GPT-series, Codex, and GLM, giving you flexibility to use your preferred provider.
- Customizable workflows: Switch between modes for planning, coding, reviewing, and debugging to match your task.
核心功能
基础AI开发能力
- ✨ 自然语言生成代码
文字描述开发需求,一键生成完整可用代码片段、函数逻辑与业务模块
- ✅ 智能自检迭代优化
AI自动自查代码漏洞、语法错误、逻辑问题,自主迭代优化代码质量
- 🧪 内置终端命令运行
编辑器内直接执行各类终端脚本,完美兼容多系统编码格式,杜绝中文乱码
- 🌐 浏览器自动化操作
支持网页流程自动化执行,轻松完成测试流程、重复操作、页面批量处理
- ⚡ 上下文行内智能补全
编码过程实时感知项目上下文,精准给出代码补全建议,提速日常编码
- 🤖 兼容最新主流AI模型
适配多款主流大模型,灵活选择适配自身使用习惯与使用场景
- 🎁 灵活密钥配置
支持免密钥本地部署使用,也可自定义配置云端接口,使用方式自由灵活
Features
Core AI & Development Capabilities
✨ Natural language to code
Describe what you need, and the agent will generate complete, working code snippets or full functions.
✅ Self-checking & iterative refinement
The AI reviews its own output, catches errors, and iteratively improves the solution for better quality.
🧪 Terminal command execution
Run shell commands directly from the chat interface, with built-in handling for cross-platform encoding (including UTF-8 and GBK/CP936) to prevent garbled output.
🌐 Browser automation
Automate UI workflows and browser interactions for testing, scraping, or repetitive tasks.
⚡ Inline autocomplete suggestions
Context-aware code completions that appear as you type, speeding up common coding patterns.
🤖 Support for latest AI models
Compatible with a wide range of modern large language models — choose the one that fits your workflow and privacy needs.
🎁 Flexible API key configuration
API keys are optional, making it easy to use with local or self-hosted model deployments.
专属增强终端功能
- 托管式虚拟终端
独立定制化内置终端,完整捕获用户输入内容与命令运行输出,便于二次开发与流程整合
- 命令执行全流程记录
自动记录命令启动、运行结束、执行日志、退出状态、执行时间等完整结构化数据,方便日志留存与流程自动化对接
- 原生命令历史快捷键
支持上下方向键快速调取历史执行命令,支持编辑修改后重新运行,使用习惯贴合主流终端
Enhanced Terminal Experience
Managed Pseudoterminal
A dedicated managed terminal that captures all user input and command output, enabling richer automation and logging workflows.
Structured command tracking
Command execution events (start, completion, output, and exit codes) are tracked, making it easier to integrate with custom logging or automation pipelines.
Familiar shell-like history
Navigate previous commands using the Up/Down arrow keys, edit them inline, and re-run them just like in a standard shell.
使用教程
- 在VS Code应用商店搜索安装「仓颉AI编程助手」,或手动安装VSIX离线安装包
- 使用VS Code打开任意项目工程文件夹
- 打开插件侧边对话面板,或通过命令面板快速唤起插件功能
- 输入自然语言指令即可生成代码、运行项目命令、完成代码调试与项目开发
常用使用场景
- 自然语言快速编写页面组件、接口函数、工具类代码
- 编辑器内直接运行启动项目、打包构建、安装依赖等终端指令
- 对已有代码进行重构优化、添加注释、编写单元测试、修复代码BUG
Getting Started
- Install the
Cangjie Code Assistant extension from the VS Code Marketplace or by installing the VSIX package.
- Open your project workspace in VS Code.
- Open the Cangjie-AI chat view or use the command palette to start using the agent.
- Use natural language prompts to generate code, run commands, or automate tasks.
Example workflow:
- Ask the agent to “Add a login form with validation to this React component”
- Run
npm run dev directly from the chat
- Request refactoring, unit tests, or bug fixes for selected code
开源协议
本项目遵循 Apache License 2.0 开源协议
您可免费使用、修改、二次开发及商用分发,使用过程中请严格遵守开源协议条款,保留相关版权声明与协议文件。
完整协议内容请访问官方协议地址:
https://www.apache.org/licenses/LICENSE-2.0
License
This project is licensed under the Apache License 2.0.
You are free to use, modify, and distribute this software (including for commercial purposes), provided that you comply with the terms of the Apache 2.0 license, including proper attribution and preservation of license notices.
For the full license text, visit:
https://www.apache.org/licenses/LICENSE-2.0
| |