This company uses MCP and has delivered 1,000 digital employees to businesses.

Business Background

Company Introduction

Junrun Human Resources Service (Shenzhen) Co., Ltd. (referred to as Junrun Human Resources) is a technology-driven comprehensive human resource solution service provider, adhering to the service philosophy of  "Customer First, Integrity and Honesty, Open and Win-Win" , relying on industry-leading technology and service capabilities, focusing on providing one-stop human resource services for the service industry, helping improve the efficiency of human resource management, while increasing the happiness index of workers and promoting social employment. As of now, the company has served over  2000  companies and more than  3  million individuals.

Digital Intelligence Layout

From 2019 to 2022, Junrun Human Resources built a one-stop digital solution covering the entire process scenarios of recruitment, social security, payroll, risk control, electronic signatures, and settlement payments. All systems are independently developed to ensure controllability and reliability of technology, and currently support more than  30  core business systems.

With the gradual decline of the demographic dividend, enterprises will also face the challenge of rising labor costs. Automating and intelligentizing a large number of repetitive transactional tasks through human-machine collaboration is the trend for future development. Since the beginning of 2023, Junrun Human Resources has begun continuous exploration in the field of AI, focusing particularly on how to construct digital employees (intelligent agents) that can deliver results for various job positions. Currently, Junrun has over  1000  digital employees working 24/7, cumulatively saving labor costs of over ten million.

Business and Technical Challenges

Internal Capability Upgrade: How to Reshape Efficiency of Existing Business through AI

Junrun Human Resources has over 30 core business systems internally, which serve as important carriers for service delivery to clients. During the continuous expansion process, a wealth of business capabilities has accumulated, which are the "skills" that digital employees need to master. Digital employees become the carriers of these "skills", and possessing these "skills" allows for better delivery of work results for various positions. For example, for digital employees in social security services, completing independent tasks such as handling social security onboarding and offboarding, verifying insured employees, responding to employee inquiries, and downloading insurance certification requires the use of existing business system capabilities.

This process essentially leverages business processes, using AI capabilities to reshape the efficiency of value delivery. Most business scenarios can consider using digital employees to achieve cost reduction and efficiency enhancement, while improving customer experience. For example:

1. In the recruitment field, through "recruitment digital employees", the entire process from writing recruitment job descriptions (JD), publishing recruitment needs on platforms, greeting candidates, introducing job information, obtaining and screening resumes to inviting for interviews can all be autonomously completed by digital employees, achieving a fully automated process and significantly improving recruitment efficiency.

2. In the social security field, through "social security digital employees", using AI + RPA technology to pay social security for 300,000 employees spread across 275 cities and performing insured verification while also responding to employee inquiries related to social security.

3. .....

For internal systems, how to leverage AI to reshape the efficiency of the core value chain and enhance business capabilities?

External Capability Integration: How to Introduce AI Innovations at Low Cost

Starting in 2023, Junrun established a special team to study ChatGPT. From the beginning with ChatGPT 3.5, to later developments such as DeepSeek, Doubao, and Tongyi Qianwen, how will the protocol layer cope with the evolution of new protocol architectures such as MCP (Model Context Protocol) and A2A (Agent-to-Agent)?

As a technology-driven human resources service company, how can Junrun quickly innovate based on the AI + human resources track?

Solutions

Q1. How can existing business be enhanced through AI?

• Unification (Technology Stack, AI Technology Foundation):

1. Introduce Alibaba's AI agent application new paradigm at Junrun.

2. Rely on the two cutting-edge model technology foundations of Tongyi and DeepSeek.

3. Based on Alibaba's Bailing (applications), Coze (workflow), and Mingyue AI + Dify (internal) intelligent development platforms.

4. Deeply customize digital workforce agents for the human resources industry.

5. Build a full-chain data protection system through Higress's security series plugins.

6. Develop while standing on the shoulders of giants, establish an agent intelligent body lifecycle management platform that conforms to Junrun's technology architecture.

• Reuse/Enhance:

1. Achieve intelligent orchestration and efficient reuse of hundreds of RESTful API capabilities across over 30 internal core systems through Higress MCP plugin.

2. Utilize Alibaba's Bailing application observability and AI plugins (AI content security, AI token rate limiting, AI statistics, logs, etc.) to ensure the usability of Junrun’s digital workforce applications and facilitate iterations.

Implementation Cases

MCP Tool (Social Security Calculator) MCP Marketplace:  

https://mcp.higress.ai/server/server10001

Q2. How can Junrun quickly innovate based on AI + human resources track?

• Innovation Model:

For the first time, proposed the "digital workforce = digital employee + flexible workforce" model, building three major capability evolution paths:

Skill Enhancement: Achieve skill integration through Higress MCP gateway configuration, achieving capability leap from "Digital Workforce (Intern) → Digital Workforce (Specialist) → Digital Workforce (Manager)".

Ecological Reuse: Integrate open-source services through MCP Marketplace and Magic Multiplication MCP Square.

On-Demand Employment: Enterprises can be activated instantly without the need for training or configuration.

From the product dimension, unify and integrate into Junrun's digital workforce platform, creating a market for digital human resource workers.

https://agent.junrunrenli.com?src=higresswx:

Why Choose Higress MCP Gateway

Technology Selection

Junrun has accumulated a large number of Agents internally. With the implementation of MCP protocols by major cloud providers, Junrun aims to build a digital workforce market, requiring Agents to have MCP capabilities to fulfill the delivery of results for digital workers.

In April 2025, Alibaba Bailing released the MCP SSE mode, directly addressing our pain points, prompting us to fully embrace and upgrade the company's AI architecture. Integrating the internal capabilities of 30+ digital systems with AI for comprehensive enhancement.

Pain Points

Pain Point 1: How to quickly convert APIs into MCP services and provide them for Agent use?

Using Higress "MCP server" plugins for on-demand one-click conversion, referencing the documentation at https://higress.cn/ai/mcp-server/, we convert parameters such as "path", "header", "query", and "body". The backend API requires no modifications; it can be configured in "MCP server" with seamless publishing of MCP services.

Pain Point 2: How to authorize MCP services for Agents?

Some of our Agents are developed on external platforms like "Alibaba Bailing". Agents need to authorize the use of MCP services. We adopt the combination of Higress open-source provided "Key Authentication" + "MCP Service" plugin for seamless onboarding. This perfectly resolves the authorization between Agents and MCP. For user permissions for Agents, we unify permission management based on our self-developed digital workforce platform.

Pain Point 3: How to expose internal network API interfaces converted into MCP services externally?

Many of our internal API services are deployed within the k8s internal network, inaccessible externally. We adjust system architecture for transformation, privatizing Higress deployment to k8s cluster. Higress communicates with internal API services through k8s DNS service names, exposing only the Higress MCP services externally. A global plan for domain names and access paths is also established.

Outlook on Higress

Enhancing user-friendliness, we hope to further improve the documentation of Higress open-source, suggesting that each plugin adds a link to usage documentation for corresponding versions. This will reduce the cost of using plugins and improve development and operation efficiency.

The Higress enterprise version offers ready-to-use capabilities with stronger performance and better stability, click here for a free trial.