zhikuncode
Codex/Claude Code/Cursor的开源增强版,专注一句话实现复杂长程任务(教育、编程、办公、生活、娱乐、游戏)。部署在你自己的服务器上,团队用浏览器打开就能编程——包括手
What is zhikuncode?
Codex/Claude Code/Cursor的开源增强版,专注一句话实现复杂长程任务(教育、编程、办公、生活、娱乐、游戏)。部署在你自己的服务器上,团队用浏览器打开就能编程——包括手机。CLI & Web UI 双入口,Multi-Agent 协作,原生直连千问/DeepSeek 等国产大模型。技能/插件/跨会话记忆,8 层安全沙箱,数据不离开你的机器。Docker 一键部署,MIT 开源,零锁定。
Overview of zhikuncode
What it is
ZhikunCode is an open-source AI programming assistant that can be deployed on your own server. It is designed as an enhanced version of Codex/Claude Code/Cursor, focusing on completing complex long-term tasks with a single instruction across domains such as education, programming, office work, daily life, entertainment, and gaming. It provides both a CLI and a Web UI, supports multi-agent collaboration, and natively connects to Chinese large language models like Qwen and DeepSeek. The system includes skills, plugins, cross-session memory, and an 8-layer security sandbox, ensuring data never leaves your machine. It can be deployed with a single Docker command and is open-sourced under the MIT license with zero vendor lock-in.
Key features
Browser-based full workflow control: deploy once, then operate from any device's browser, including mobile phones, without installing a client. Dual-view task workspace: switch between a concise workspace and a development workspace within the same session, changing only information organization without affecting backend execution. Agent delegation and background tasks: five sub-agent roles with independent contexts and parallel execution, with synchronous result return and queryable, cancellable background tasks. Unified authorization security architecture: all core tools pass through a Tool Gateway with normalized input, risk analysis, invariant checks, grant matching, dynamic re-check before execution, and structured result auditing. Native support for Chinese LLMs: out-of-the-box connectivity to Qwen, DeepSeek, Moonshot, Zhipu GLM, and MiniMax via domestic networks without requiring VPN. Docker one-click deployment: docker compose up -d starts the Java backend and built-in static frontend, with data stored locally. Smart context management: six-level compression cascade, incremental folding, precise token counting, self-correction loop, and image context governance for seamless long conversations.
Use cases
ZhikunCode is suitable for a wide range of tasks including educational content creation, software development, office automation, daily life assistance, entertainment, and game development. It can be used by teams to collaboratively code from a browser, handle complex multi-step workflows, and perform code review and comparison. The system is designed for scenarios requiring long-running, multi-agent collaboration with strict security and data privacy requirements.
Getting started
docker compose up -d
License and pricing
This project is open-sourced under the MIT license. The pricing type is open source, meaning it is free to use and modify.
Based on the project's public README and repository metadata.