Why it matters
This release provides developers with a vast array of specialized AI agents, enabling more sophisticated and context-aware automation for tasks across various business functions. The inclusion of China-specific agents and broad tool support enhances its applicability for global and local development teams.

What changed

The agency-agents-zh GitHub repository has been updated to feature 328 AI expert roles, an increase from previous versions. This collection includes 261 translated roles from the English upstream project and introduces 67 original agents tailored for the Chinese market. These new agents cater to platforms and sectors such as Xiaohongshu, Douyin, WeChat, Bilibili, Feishu, DingTalk, cross-border e-commerce, government-to-government (ToG) services, medical compliance, Qt industrial upper computers, mechanical design, and livestock and poultry farming record verification. The project supports integration with 20 different AI programming tools, including Claude Code, Cursor, and Copilot, and covers 22 business departments, ranging from CEO/CTO/CMO roles to engineering, design, marketing, product, gaming, security, GIS, and finance.

A notable addition is the 'engram' feature, designed to provide memory persistence for AI coding tools like Claude Code, Codex, Cursor, and Gemini CLI. Engram allows these tools to remember past conversations, file modifications, and solutions, automatically recalling this context in subsequent sessions. This memory capability is designed to be cross-tool compatible, meaning context established with one tool can be accessed by another. The system operates entirely locally with zero dependencies and offers Chinese language support.

The project also offers an accompanying orchestration tool, Agency Orchestrator, which allows multiple AI experts to collaborate automatically based on a Directed Acyclic Graph (DAG). Users can initiate complex tasks with a single sentence. A desktop client is available for macOS, Windows, and Linux, alongside an online experience at ao.aiolaola.com/experts.

Installation and Integration

Installation into AI coding tools can be done via a shell script (./scripts/install.sh), which automatically detects installed tools or allows for specific tool selection. For tools like OpenClaw, a conversion script (./scripts/convert.sh) is first required to transform agent definitions into a compatible format. Manual copying of agent definition files (e.g., SOUL.md, AGENTS.md, IDENTITY.md) is also supported for specific tools like Claude Code and GitHub Copilot. Additionally, the role library can be accessed via npm (npm install agency-agents-zh).

Agent Categories

The agents are organized into broad categories, including "Company Management" (CEO, CTO, CMO, CFO, etc.) and "Engineering," which itself encompasses a wide range of specializations such as Frontend Developer, Backend Architect, AI Engineer, DevOps Automation Engineer, Security Engineer, and many more niche roles like Embedded Algorithm Engineer, Mechanical Design Engineer, and RAG Pipeline Engineer.

Why it matters for builders

This extensive release significantly expands the toolkit available to AI builders by providing a comprehensive library of specialized AI agents. The granular definition of roles, workflows, and deliverables for each agent moves beyond generic prompts, enabling more precise and effective AI-assisted development. The inclusion of China-specific agents and broad compatibility with popular coding tools makes this resource highly adaptable for diverse project needs and team structures.

Practical impact

Developers can leverage these agents to automate complex workflows, from strategic business planning to detailed engineering tasks. For instance, a team could orchestrate a "management meeting" by assigning roles like CEO, CTO, and CMO to analyze a single issue from their respective perspectives. The 'engram' feature offers a practical solution for maintaining context in AI coding sessions, reducing repetitive input and improving efficiency. Builders can explore the online expert library, install agents into their preferred coding environments, or use the Agency Orchestrator for multi-agent collaboration.

Caveats and source limits

The provided source primarily details the features and scope of the agency-agents-zh GitHub repository. Specific performance benchmarks for the agents or the 'engram' feature are not included. While the repository lists support for 20 AI programming tools, the exact integration depth and potential limitations for each tool are not elaborated upon. Information regarding the development roadmap, future updates, or community contribution guidelines beyond the provided installation methods is also absent. The source mentions a desktop client and an online experience, but detailed usage instructions or feature sets for these platforms are not fully described.

Sources

Written with AI assistance from the linked sources; every claim below was checked against them automatically. How we produce articles.

Claim check: 8/8 supported claims - 8 evidence links - 100% avg confidence
  • The agency-agents-zh project offers 328 plug-and-play AI expert roles.supported - github.com
  • 67 of the AI expert roles are original agents tailored for the Chinese market.supported - github.com
  • The agents cover 22 business departments.supported - github.com
  • The project supports integration with 20 AI programming tools, including Claude Code, Cursor, and Copilot.supported - github.com
  • The 'engram' feature provides memory persistence for AI coding tools like Claude Code, Codex, Cursor, and Gemini CLI.supported - github.com
  • Engram operates entirely locally with zero dependencies and supports Chinese language.supported - github.com
  • Agency Orchestrator allows multiple AI experts to collaborate automatically based on a DAG.supported - github.com
  • A desktop client for Agency Orchestrator is available for macOS, Windows, and Linux.supported - github.com

Caveats

  • Single-source caution: verify critical details at the linked source.
Radar score 86/100 - how it was calculated
Reliability82
Freshness92
Novelty71
Technical89
Developer96
Ecosystem66
Confidence100
  • Reliability 82: GitHub metadata supports source trust
  • Freshness 92: Fresh GitHub activity
  • Novelty 71: Novelty blends source metadata and enrichment
  • Technical 89: Repository technical metadata
  • Developer 96: Developer tooling signals
  • Ecosystem 66: Developer-oriented GitHub signal
  • Confidence 100: Claims have reliable evidence
Share
XLinkedInHacker News

Discussion

Loading comments...

Related articles

Agents - Nov 17, 2026BuilderIO Releases Agent-Native TypeScript FrameworkBuilderIO has released Agent-Native, an open-source TypeScript framework for building agentic applications. It enables agents to pair autonomous work with purpose-built UIs, allowing shared actions, data, and application state between the agent and the user interface.Agents - Nov 18, 2026faf-cli v8.10.0 Enhances AI Agent Context with Project FactsThe faf-cli tool has released version 8.10.0, introducing "Project Facts" to provide AI coding agents with versioned, stack-specific context. This update ensures that AI agents receive accurate information about a project's real stack, reducing guesswork and improving code generation.Agents - Nov 19, 2026ntfy-mcp-server: Manage ntfy Notifications via MCPThe ntfy-mcp-server project provides tools to send, manage, and replay ntfy push notifications using the Model Context Protocol (MCP). It offers four distinct tools for publishing messages, managing existing ones, fetching message history, and searching for emoji tags.Agents - Nov 17, 2026Overleaf CLI (olcli) v0.17.0 Adds AI Agent IntegrationThe Overleaf CLI tool, olcli, has released version 0.17.0, introducing an MCP server for AI agents. This allows AI agents to interact with Overleaf projects, prompting users before making changes.Agents - Nov 17, 2026Trustabl Agent Dashboard for AI Reliability ScansTrustabl has released Agent Dashboard, a local web UI for running reliability scans, skill tests, and network access probes on AI agents. It integrates Trustabl's Analyzer, Forge, and Probe tools, simplifying agent development and testing.Agents - Nov 20, 2026PraisonAI v4.7.26: Offline AI Agents & Enterprise HandoverPraisonAI has released version 4.7.26 of its open-source Python agent framework, enabling fully offline AI agent execution. The update also introduces an early access feature for highly regulated teams, allowing AI tasks to be handed over to self-hosted models with a verifiable audit trail.