Unlock JD’s Open‑Source JoyAgent: Build, Deploy, and Customize AI Agents

For the past 25 years, the rise of RAG knowledge bases and MCP services has sparked a wave of AI agent design across many internet companies.

Programmers were among the first to use various AI agent products such as Coze, Dify, Claude, Cursor, and Trae.ai, and naturally wonder how these agents are implemented.

Fortunately, the first wave of enterprise‑grade open‑source AI agent code has arrived: joyagent‑jdgenie .

The open‑source JoyAgent‑JDGenie is a general, lightweight, and extensible multi‑agent framework that can be used out‑of‑the‑box. It offers Docker deployment, local script deployment, and an IntelliJ IDEA project that runs as a SpringBoot application for debugging.

The project includes Java, Python, and TypeScript components, with Java as the primary language (about 0.57 k lines of core code), making it easy for developers to start learning AI agent development.

JoyAgent‑JDGenie provides front‑end, back‑end, framework, engine, and core sub‑agents such as report‑generation, code, PPT, and file agents. Users can mount additional sub‑agents or tools to address new scenarios. In the GAIA validation set the framework achieved 75.15 % accuracy and 65.12 % on the test set, surpassing well‑known products like OWL (CAMEL), Smolagent, LRC‑Huawei, xManus, and AutoAgent.

Beyond the metrics, the framework uses standard protocols and a lightweight design, acting as a universal foundation that can be extended to various services.

1. Create a poster cover

A backend engineer with 10+ years of experience used JoyAgent to generate a side‑project poster for inviting collaborators. The AI agent iteratively refined the design until satisfactory, demonstrating how AI agents can handle design and UI tasks.

Website: https://joyagent-genie.jdcloud.com/

2. Download, deploy, and run locally

The source code is available at GitHub . The following steps outline a typical local setup.

2.1 Configuration

Copy the .env file under genie-tool and edit the configuration. Adjust project settings as needed.

Also modify the project configuration to enable local testing.

The execution plan is a composite prompt that drives the AI agent’s automatic analysis and task execution.

2.2 Build the Docker image

Run the command: docker build -t genie:latest . Note that the first build may take a long time and might need to be retried.

2.3 Start the container

Execute:

docker run -d -p 3000:3000 -p 8081:8081 -p 8188:8188 -p 1601:1601 \
-e OPENAI_BASE_URL="https://***.cn/v1" \
-e OPENAI_API_KEY="sk-...your_key..." \
--name genie-app genie:latest

Replace OPENAI_BASE_URL and OPENAI_API_KEY with your own values. Adjust port mapping if necessary (e.g., 8080:8080 for local service debugging).

2.4 Access the service

With JDK 17 installed, start the backend and open http://localhost:3000/ in a browser. You can now interact with the AI agent.

JoyAgent‑JDGenie is a lightweight yet powerful AI agent platform suitable for enterprise deployment and iterative development. Its clear code structure makes it easy to extend, and a one‑click Docker deployment script would further lower the entry barrier for newcomers.