zhikunqingtao/zhikuncode

What it solves

ZhikunCode is an open-source AI programming assistant designed to automate the full software development lifecycle. It solves the problem of fragmented AI coding tools by providing a self-hosted, browser-based environment where multiple AI agents can collaborate to write, test, and verify code across various platforms (including mobile devices) without requiring a local client installation.

How it works

The system uses a multi-agent architecture (supporting Team, Swarm, and SubAgent modes) to decompose complex tasks. It is built with a Java Spring Boot backend, a React frontend, and a Python capability service for code analysis. It integrates with various LLMs (including domestic Chinese models like Qwen, DeepSeek, and Kimi) and features a sophisticated "Context Cascade" system to manage long conversations through six levels of compression. It also includes a runtime verification framework to ensure the generated code actually works via browser, HTTP API, or automated tests.

Who it’s for

It is aimed at developers who want a powerful, self-hosted AI coding assistant that can handle complex engineering tasks, as well as those who prefer using domestic Chinese LLMs with direct connectivity.

Highlights

• Browser-Based Control: Full development workflow accessible via any browser, including mobile and iPad.
• Multi-Agent Collaboration: Supports fixed roles (Team), dynamic negotiation (Swarm), and master-slave delegation (SubAgent).
• Advanced Context Management: A six-layer compression cascade to handle ultra-long dialogues without losing critical information.
• Runtime Verification: A built-in framework that generates evidence (screenshots, logs, diffs) to verify the correctness of the output.
• Domestic LLM Integration: Out-of-the-box support for Qwen, DeepSeek, Moonshot, and ZhipuGLM.
• Docker Deployment: Easy one-click setup using Docker Compose.