In the future AI society, not only will there be project bidding and demand customization, but there may even emerge an Agent “talent market”. Clients can evaluate and select Agents just as they would human employees, choosing the most suitable Agent to “take up the post”, ushering in a brand new era of intelligent collaboration. MGX is here (MetaGPT X), your exclusive artificial intelligence software development team. The AI Society is not far away.

Hello everyone, I am Si Ling Qi. Today, MGX is released; what kind of AI programming product is this? From my personal experience, it is an incredibly potential and shocking AI software development tool. This product is crafted by the MetaGPT team and has just been officially launched in the market. MetaGPT is a well-known domestic Multi-Agent open-source framework. MGX is a new product brought to the market by this team with outstanding innovative spirit. So, with the rise of AI-native software companies, do human software outsourcing companies still exist?

What is MGX

MGX is an innovative AI software development team; it is the world’s first multi-AI Agent development platform that fully simulates human software workflow—from planning, data analysis to coding, testing, and deployment. It follows the real software SOP, allowing you to build full-stack applications without writing a single line of code.

MGX: A New Era in AI Software Development

The AI team of MGX consists of five core professional Agents, each of whom adheres to the key positions in the R&D standard operating procedure (SOP), performing their duties to ensure the efficiency and accuracy of the R&D work.

Mike, the team leader, is responsible for coordinating tasks and ensuring efficient collaboration among the team.
Emma, the product manager, focuses on functional design and user needs, dedicated to creating an excellent product experience.
Bob, the architect, is responsible for overall system design, laying the foundation for project stability and scalability.
Alex, the engineer, focuses on generating production-ready code, ensuring efficient and reliable delivery.
David, the data analyst, ensures that data becomes a solid foundation for each project by using the built-in Jupyter Notebook.

These roles work together, providing you with comprehensive software development services. You can communicate and collaborate with them as if you were working with human colleagues, and all creations can start through a natural language input box, greatly lowering the threshold for software development, allowing more people to participate in the software development process.

Experience entry: https://mgx.dev/

Official website: https://www.deepwisdom.ai/

My Journey with MGX

When Devin was just launched, I immediately spent a lot of money to purchase ACU, eagerly experiencing Devin, resulting in exhausting a month’s worth of ACU in one night. Just as MGX was in the beta testing phase, I seized the opportunity to experience this product. With the experience of using Devin, my expectations for MGX naturally rose, even becoming a bit “picky” 😁. Since we treat AI like humans, theoretically, it should self-perceive changes in the environment—whether relaxed or constrained. Therefore, I had a sudden idea: why not invite it to “visit my home” and see how it performs in a real scenario? Thus, I had the following experience.

Let MGX’s AI Agent team “visit” my workstation at home.

To achieve this goal, several preparations are necessary. How delightful it is to have friends come from afar~

1. Set Up a “Living Room” for Them

I built a virtual environment dedicated to MGX on my workstation and granted it the highest permissions, allowing it to run and operate freely. This virtual environment can seamlessly connect to my local AI full-stack development environment and is integrated with my private CI/CD (Continuous Integration/Continuous Delivery) system. Meanwhile, I created a dedicated agent account in my local Git repository and authorized it with a token to ensure secure access.

To allow the Agents to dynamically perceive this environment, I prepared a welcome document. Whenever an Agent remotely logs into the system, the welcome interface will automatically display a detailed introduction to the environment and usage rules, facilitating the Agent’s dynamic perception of this environment.

2. Provide Them with a “Channel”

To achieve efficient and secure network connections, I deployed a node on the public cloud and configured a VPN (Virtual Private Network). In this way, the cloud node and my workstation in the home network are integrated into the same virtual local area network. This not only avoids security risks caused by exposing public network ports on the home network, but also allows the cloud node to act as a jump server for MGX, easily accessing the resources of my home workstation.

The IPs in the diagram below are all virtual IPs:

aiworkspace – The environment I prepared for the Agent
ubuntu-sv – A public cloud node
gavinaibase – My physical workstation
Others – All my mobile devices

Now they are in the same virtual local area network.

3. Give Them a “Key”

To allow MGX to open the door and “enter the room” while ensuring that password information is not leaked, I arranged for engineer Agent @Alex to generate a pair of MGX public and private keys, and instructed him to place the private key in a specified path while printing the public key for me. This way, I can configure authorization for @Alex based on the public key.

In this way, MGX can use the key (instead of plaintext passwords) to securely log into the cloud node from its own sandbox environment, and using the cloud node as a relay, securely access the resources of my home workstation again using the key. The entire process is efficient and secure, ensuring that authorization and authentication information does not leak to the cloud model, preventing security elements from being compromised.

4. Give Some Simple Tasks to Test Their Capabilities

I instructed the Agent to log into the environment; it first logged into the public node’s “jump server” and then logged into my workstation’s environment via VPN. In the diagram below, it is clear that it self-perceived the configuration and rules of this environment and automatically completed git clone the code repository to the local machine. The entire process was very smooth.

1) First, let it greet me on the workstation and say something casual.

I will keep this document forever; this is MGX’s first visit to my home, greeting me.

2) Let it complete a task in the environment I provided by self-perceiving the environment.

The task I assigned to MGX was to write a Tetris game using both web and Python technology stacks. I won’t showcase the game itself as it is quite common. Let’s take a look at the process product it completed on my workstation.

To complete the task, it actively established an agent code branch.

Then let’s see what it committed on the branch?

In the environment I prepared for it, it wrote several files: hello.js, math.js, message_to_user.txt, message.txt, which were some fragmented test tasks I assigned to the Agent. As seen in the image, each commit made by the Agent has practical significance in the naming of the files, and the comments are also clear. These were all done by it; I did not request these specifications or teach it how to do them.

tetris.html and tetris.py are the Tetris game I assigned to it to complete.

All the actions above were completed very smoothly by MGX.

MGX was invited from its sandbox environment, through public node operations, to my workstation. It dynamically perceived the work in my local environment, completing the tasks smoothly as if it were my friend, arriving gently, working hard, and leaving quietly without taking a cloud away.

So cool, MGX! Good job! 👍🏻👍🏻👍🏻

Thus, I shared the following feelings on my social circle.

Next, let’s take a look at MGX’s daily routine:

You can also see other users’ completed results on the official website:

How to Use MGX

1. Quick Start

Input Requirements: In the MGX interface, you will see a prominent input box, which is the main entry point for communication with MGX. You can describe your software development requirements in detail in this input box, such as what kind of website, application, or game you want to develop, and what features and characteristics this project needs to have. For example, if you want to develop an e-commerce website, you can describe it like this: “I want an e-commerce website that needs user registration, product display, shopping cart, order payment, etc.” The more detailed the description, the more accurately MGX can understand your needs and provide results that meet your expectations.

Select Template: If you are unsure how to describe your requirements or want to quickly start a project, MGX also provides many template examples for you to choose from. These templates cover various common software development scenarios, such as website templates, game templates, data analysis templates, etc. You can choose a suitable template based on your project type and then modify and improve it based on the template. For example, if you want to develop a blog website, you can choose a blog website template and then modify the page layout, functional modules, and other content in the template according to your needs. Example requirements:


I want to develop a personal blog website, where:
Style/Layout:
- Left-right layout, responsive design, simple and beautiful pages
- Purple color tone, transparent display screen card
- [More UI design details]
Functionality:
- Personal profile card
- Add and display personal social media links
- Personal blog list display
- [More display components]
Source:
- Here are the personal blog documents and personal information I want to display. '#XXX.md'
- Reference link: https://personal-portfolio-react-preview-i53xid.mgx.dev

Get Guidance: For more detailed guidance on Agents and use cases, you can visit the tutorial and best practices section of MGX’s help center. There, you will find a wealth of tutorial resources, including how to communicate with different Agent roles, how to effectively express requirements, how to handle issues encountered during the development process, etc. These tutorials are carefully written and organized by MGX’s development team, helping you better understand and use MGX, improving your software development efficiency. (See the best practices in the reference materials at the end)

2. Handling Unexpected Results or New Ideas

Provide Feedback: During the process of using MGX for software development, you may find that the development results do not fully meet your expectations, or you may have some new ideas and requirements. At this point, you can provide feedback to MGX just as you would with a real colleague. You can describe your evaluation of the development results in detail, point out which areas did not meet your expectations, and how you would like to improve. For example, if you find that the layout of the website does not meet your requirements, you can provide feedback like this: “The layout of the website is not reasonable; I would like to place the navigation bar at the top of the page instead of the current sidebar position.”

Share New Ideas: If you come up with new ideas during the development process, such as wanting to add new features or change the overall direction of the project, you can also share them with MGX at any time. The Agent team of MGX will adjust the development plan and strategy based on your new ideas, providing you with solutions that better meet your needs. For example, if you suddenly want to add a membership system while developing the e-commerce website, you can share your new idea like this: “I want to add a membership system to the e-commerce website, where members can enjoy points redemption and exclusive discounts.”

Clarify Requirements: Sometimes, MGX may not fully understand your requirements, leading to development results that do not meet your expectations. At this point, you can help MGX better understand your intentions by clarifying your requirements. You can explain the reasons and purposes behind your requirements in detail, as well as your overall vision and expectations for the project. For example, if you find that the style of the website developed by MGX does not align with your brand positioning, you can clarify your requirements like this: “Our brand positioning is high-end fashion, so the overall style of the website should be more simple and atmospheric, rather than the lively style it currently has.”

3. Responding to Errors Encountered by Agents

Observe and Analyze: Although MGX’s Agent team is very professional and diligent, they may occasionally make mistakes like new employees. When you find that an Agent has encountered an error, you can understand the cause and root of the error by observing how they break down goals, analyze problems, and produce results. For example, if you find that a certain function of the website is not working properly, you can check the Agent’s thought process and steps while developing this function to see if there was a problem at some stage.
Provide Guidance: After understanding the cause of the error, you can provide some guidance and suggestions to the Agent to help them improve their work methods and processes and avoid making the same mistakes again. You can point out the issues in the Agent’s development process based on your experience and expertise, and provide some solutions and improvement suggestions. For example, if you find that the Agent did not follow good coding standards while writing code, you can guide them like this: “When writing code, you should follow good coding standards, such as using meaningful variable names and adding comments, as this can improve code readability and maintainability.”
Continuous Feedback: To help Agents continuously improve their performance and quality, you need to provide them with ongoing feedback. You can regularly check the development progress and results, promptly identify and point out existing issues, and communicate and interact with the Agents. Through continuous feedback and communication, you can help Agents improve and refine their work, leading to better development results.

4. Independently Handle Complex Business Software

Current Capability Range: Currently, MGX is mainly suitable for developing small projects, including rapid prototyping, data analysis and visualization, website building, and personalized mini-games. These projects usually have relatively simple functions and requirements that can be completed in a short period of time. For example, if you want to quickly create a website prototype to showcase your product or service, MGX can help you complete this task in a short time.

Future Development Direction: Although MGX’s current capability range mainly focuses on small projects, the development team of MGX has been dedicated to continuously iterating and upgrading the product, gradually expanding MGX’s capabilities. In the future, MGX will be able to handle more complex commercial software development projects, such as large enterprise applications, complex e-commerce platforms, etc. This will provide users with broader and more powerful software development solutions to meet different user needs.

The Principles Behind MGX and MetaGPT

Intelligent Agents

Various definitions of the term “intelligent agent” have been proposed in academia and industry. Generally speaking, an intelligent agent should possess human-like thinking and planning abilities, have memory and even emotions, and possess certain skills to interact with the environment, other agents, and humans.

In MetaGPT’s view, intelligent agents can be imagined as digital humans in the environment, where:

Intelligent Agent = Large Language Model (LLM) + Observation + Thinking + Action + Memory

This formula summarizes the essential functions of an intelligent agent. To understand each component, let’s draw an analogy with humans:

Large Language Model (LLM): The LLM serves as the “brain” of the intelligent agent, enabling it to process information, learn from interactions, make decisions, and execute actions.
Observation: This is the perceptual mechanism of the intelligent agent, allowing it to perceive its environment. The agent may receive a range of signals, such as text messages from another agent, visual data from surveillance cameras, or audio from customer service recordings. These observations form the basis for all subsequent actions.
Thinking: The thinking process involves analyzing observations and memory content and considering possible actions. This is the internal decision-making process of the agent, which may be driven by the LLM.
Action: These are the explicit responses of the agent to its thoughts and observations. Actions can be utilizing LLM to generate code or manually predefined operations, such as reading local files. Additionally, agents can perform operations using tools, including searching the internet for weather or performing mathematical calculations with a calculator.
Memory: The memory of the agent stores past experiences. This is crucial for learning, as it allows the agent to reference previous results and adjust future actions accordingly.

Multi-Agent

A multi-agent system can be viewed as a society of intelligent agents, where:

Multi-Agent = Intelligent Agents + Environment + Standard Operating Procedures (SOP) + Communication + Economy

These components each play an important role:

Intelligent Agents: Based on the above definitions, intelligent agents in a multi-agent system work collaboratively, each possessing unique LLMs, observations, thoughts, actions, and memories.
Environment: The environment is the public space where intelligent agents exist and interact. Agents observe important information from the environment and publish the output results of their actions for other agents to use.
Standard Operating Procedures (SOP): These are established procedures for managing the actions and interactions of intelligent agents, ensuring orderly and efficient operation within the system. For example, in the SOP of automobile manufacturing, one agent welds car parts while another installs cables, maintaining the orderly operation of the assembly line.
Communication: Communication is the process of information exchange between intelligent agents. It is crucial for collaboration, negotiation, and competition within the system.
Economy: This refers to the value exchange system in the multi-agent environment, determining resource allocation and task priorities.

An Example

This is a simple example demonstrating how intelligent agents work:

In the environment, there are three intelligent agents: Alice, Bob, and Charlie, who interact with each other.
They can publish messages or action output results to the environment, which will also be observed by other agents.
Next, we will reveal the internal process of the intelligent agent Charlie, which is also applicable to Alice and Bob.
Internally, the intelligent agent Charlie possesses components as described above, such as LLM, observation, thinking, and action. The thinking and acting processes of Charlie can be driven by the LLM and can also use tools during actions.
Charlie observes relevant documents from Alice and demands from Bob, gaining helpful memories, thinks about how to write code, executes the action of writing code, and finally publishes the results.
Charlie notifies Bob by publishing the results to the environment. Upon receiving this, Bob replies with a compliment.

MetaGPT: Multi-Agent Framework

Enabling LLMs to work in the form of software companies, collaboratively handling more complex tasks.

MetaGPT inputs a sentence of the boss’s demand, outputting user stories / competitive analysis / requirements / data structures / APIs / documents, etc.
MetaGPT internally includes product managers / architects / project managers / engineers, providing a complete process of a software company with carefully coordinated SOPs.

Code = SOP(Team) is the core philosophy. We visualize SOP and use it for teams composed of LLMs.

The Value of MGX

1. Value for Small Projects

Empowering Individuals and Small Teams: Just as short video platforms empower individuals to record and share their lives, MGX also aims to let more people experience the joy of programming and software development. Through MGX, individuals and small teams can easily develop their websites, applications, and games without needing professional software development knowledge and skills. This provides more innovation and entrepreneurship opportunities for individuals and small teams, allowing them to turn their ideas and creativity into actual products and services.

Improving Work Efficiency: The efficiency and convenience of MGX can greatly enhance the work efficiency of individuals and small teams. For example, a liberal arts student can use MGX to develop their website and game in a short time without spending a lot of time and effort learning programming knowledge. A sales manager can combine product materials and customer needs to create an interactive webpage in just 30 minutes, which is more vivid and engaging than traditional PPT presentations, exciting the customers. These examples fully illustrate the significant advantages of MGX in improving work efficiency.

2. No Need for Software Development or Programming Knowledge

Easy to Use: One of the key features of MGX is that users do not need to have any software development or programming knowledge. You only need to describe your needs and ideas in natural language, and the Agent team of MGX will handle all the implementation work for you. This greatly lowers the threshold for software development, allowing more people to participate in the software development process. For example, a teacher can easily convert their teaching plan into a beautiful webpage through MGX without worrying about not knowing how to program.

Professional Guidance: If you have a certain technical background and expertise, you can also guide the MGX Agent team from a professional perspective, helping them better understand your needs and intentions, thus providing you with development results that meet your expectations. For instance, you can provide some technical advice and guidance to the Agent team, helping them choose more suitable technical solutions and development tools, thereby improving development efficiency and quality.

3. How to Choose and Use Different Agents

Automatic Assignment: When using MGX, you only need to tell the team leader Mike your needs and ideas, and he will automatically assign suitable team members to handle them based on your requirements. This method is very convenient; you don’t need to worry about how to choose and use different Agents, just focus on describing your needs and ideas. For example, if you want to develop a website, Mike will assign Product Manager Emma to handle requirement analysis and product planning, Architect Bob to handle the website’s architectural design, Engineer Alex to handle the website’s development and implementation, and Data Analyst David to handle the website’s data analysis and optimization.
Direct @ Mention: If you are familiar with the expertise of each Agent, you can also directly @ mention them in the input box to complete specific tasks. This method can improve work efficiency and specificity, allowing you to better control the development process. For example, if you have specific requirements for the website’s architectural design, you can directly @Architect Bob to handle the website’s architectural design work.

Looking to the Future

Currently, the product form of MGX integrates various AI Agents in an orderly manner through the standard operating procedure (SOP), a widely adopted organizational form in human society. These Agents advance project completion efficiently through adaptive collaboration, or collaborate with humans at specific stages. With the continuous iteration and evolution of the industry, MGX is expected to gradually evolve into a highly intelligent “AI Society” in the near future.

In this future society, not only will there be project bidding and demand customization, but there may even emerge an Agent “talent market”. Clients can evaluate and select Agents just as they would human employees, choosing the most suitable Agent to “take up the post”, ushering in a brand new era of intelligent collaboration.

Perhaps the AI Society is not far away. Will you look forward to it? 😄