Agency-Agents Chinese Version: Ready-to-Use AI Agent Library 188

What is Agency-Agents

Agency-Agents Chinese version (agency-agents-zh) is the Chinese community edition of the original English agency-agents, maintained by jnMetaCode. It is a plug-and-play library of AI expert roles, where each agent has its own persona, professional workflow, and deliverables—it's not a generic prompt template. The project is positioned as "Your AI Dream Team," covering business scenarios ranging from front-end development to Xiaohongshu operations.

Open source address: https://github.com/jnMetaCode/agency-agents-zh

Features

• 211 AI expert roles: covering 18 departments including engineering, design, marketing, product, gaming, security, and finance

• Bilingual content composition: 165 English translations + 46 original Chinese market agents

• Multi-tool compatibility: supports 16 AI programming tools including Claude Code, Cursor, GitHub Copilot, OpenClaw, Trae, Qwen Code, Gemini CLI, etc.

• Deep localization: customized dedicated agents for Chinese platform ecosystems such as Xiaohongshu, Douyin, WeChat, Bilibili, Feishu, and DingTalk

• Multi-agent collaboration: can work with the agency-orchestrator orchestration engine to define workflows via YAML for automatic multi-role collaboration

Use Cases

• Going-global product MVP: combine roles such as front-end developer, back-end architect, growth hacker, and rapid prototyper to iterate on products quickly.

• Xiaohongshu brand promotion: coordinate the Xiaohongshu operations expert, content creator, brand guardian, and data analysis expert to complete the full process from seeding to conversion.

• Cross-border e-commerce operations: use cross-border e-commerce category experts, inventory forecasting experts, and logistics route optimizers for end-to-end management.

• Enterprise compliance and risk control: invoke contract review experts, invoice management experts, and financial risk control analysts to handle legal and financial affairs.

• AI programming assistance: load specific expert personas in tools like Cursor or Claude Code to improve code quality and development efficiency.

Agency-Agents Pros and Cons

Pros

• Out-of-the-box, supports one-click installation to mainstream AI programming tools

• Complete role definitions, including four dimensions: identity, rules, workflow, and deliverables

• Highly localized, covering mainstream Chinese internet platforms

• Open source under MIT license, free for commercial use

• Active community with QQ group, WeChat group, and WeChat public account support

Cons

• Some tools (like Cursor) have too many rule files when fully installed, which may affect matching accuracy; it's recommended to pick 10-20 commonly used roles

• Role quality varies; some translated agents may need fine-tuning based on actual business needs

• Collaboration with agency-orchestrator is an advanced usage and has a certain learning curve for beginners

How to Use Agency-Agents

Installation

• Clone the repository:

  git clone https://github.com/jnMetaCode/agency-agents-zh

• Choose an installation method:

• Claude Code: copy directly to ~/.claude/agents/

• Cursor: run ./scripts/convert.sh --tool cursor, install to project .cursor/rules/

• OpenClaw (recommended): automatically split into three files—SOUL.md, AGENTS.md, IDENTITY.md—natively supporting multi-agent collaboration

• Activate and use: activate via natural language in the AI tool, e.g., "Activate front-end developer mode, help me build a React component."

Practical Examples

You can assemble a virtual project team on demand to handle complex tasks:

• Develop an e-commerce mini-program MVP: product manager outputs PRD → front-end developer builds pages → back-end architect designs APIs → security engineer reviews logic → reality checker validates before launch.

• Plan a Xiaohongshu brand marketing campaign: Xiaohongshu operations expert defines topics → content creator writes seeding copy → visual storyteller optimizes images → community manager maintains the comment section.

⚠️ Usage Recommendations

• Load on demand: for first-time use, it's recommended to install only 3-5 of your most commonly used roles to avoid causing interference in the tool.

• Manual tuning: you can directly open the Markdown files in the agents/ directory and modify the "workflow" or "deliverable template" to suit your personal usage habits.

FAQ

Q: What are the differences from the original English version?

A: Based on a complete translation, the Chinese version adds 46 original Chinese market agents and is deeply adapted to domestic platform ecosystems.

Q: Is commercial use supported?

A: MIT license, free for commercial or personal projects, no attribution required.

Q: How can I contribute?

A: You can translate agents, improve content, or add new Chinese platform agents. See the project's CONTRIBUTING.md for details.

Q: Is there a配套 multi-agent orchestration tool?

A: Yes, there is a sister project agency-orchestrator, which supports orchestrating automatic collaboration of 211 roles via YAML and is compatible with 9 LLMs including DeepSeek, Claude, and OpenAI.

Review

This project has gained nearly 3,000 Stars on GitHub and is a benchmark case in the practice of AI agent role-based design. Its core value lies in upgrading "prompt engineering" to "persona engineering"—each agent defines not only what to do, but also how to do it, what to deliver, and how to communicate. Compared with ordinary prompts, these roles have complete identity memory, communication style, and workflows, which can significantly improve the professionalism and consistency of AI output.

The original Chinese market content is the biggest highlight. The 46 local agents fill the gaps in overseas projects for domestic platform operations, government compliance, e-commerce ecosystems, and other scenarios, making it very practical for Chinese users.