The Clawdbot has become extremely popular, and I see the boundless potential path of Tencent Gold Coins.

There is a project on GitHub that skyrocketed by 9,000 stars in just one day, soaring from 7.9K in the morning to over 17K in the evening. As of the time of writing, this project has exceeded 40,000 stars.

This kind of growth rate is extremely rare in open - source projects. Moreover, since this project can run on Apple's M4 chips, it has directly triggered a rush to buy Mac minis.

And this project is Clawdbot.

Over the weekend, the entire Silicon Valley AI circle was completely engrossed in Clawdbot.

Andrej Karpathy, the former head of AI at Tesla, publicly gave his thumbs - up. Logan Kilpatrick, the head of Google's AI products, followed suit and bought a Mac mini. Federico Viticci of MacStories burned through 180 million Anthropic API Tokens in just one week just to experience Clawdbot.

Seeing the popularity of Clawdbot, I suddenly realized: Isn't this the replicable opportunity for Yuanbao in the WeChat ecosystem?

But before exploring the possibilities for Yuanbao, let's first understand what Clawdbot is and why it has conquered so many AI experts in just a few days.

01

What is Clawdbot?

In a nutshell, Clawdbot is an open - source, self - hostable AI assistant.

Its core innovation lies in the fact that you don't need to open a dedicated website or app. You can simply send a message through your daily communication apps (such as WhatsApp, Telegram, iMessage, Discord, Slack, etc.) to command it to complete various tasks.

It's like having an employee who never takes a break sitting in front of your computer. When you get off work and want to watch a TV show, you can tell Clawdbot your needs, and it will open the corresponding page. By the time you get home and sit in front of your computer, you can directly watch the TV show you want.

Clawdbot's design logic is that when you send an instruction through a messaging app, it will call the large - language model running in the background, convert your needs into local Shell scripts, and execute them on your computer.

In other words, it doesn't just tell you how to do something; it does it for you directly.

Its workflow follows the architecture of "Messaging App ⇄ Gateway ⇄ AI Model + Tools". The gateway, which serves as the coordinating center, runs on your own device and is responsible for message routing, state management, scheduled tasks, and service integration.

However, here comes the dilemma. To make the AI more powerful and useful, it needs extremely high permissions (such as reading and writing files, executing terminal commands). But for security reasons, these permissions must be restricted.

Clawdbot's solution is to introduce a "session isolation" mechanism.

It divides usage scenarios into two categories: one is the "main session", which is when you chat with it one - on - one; the other is the "non - main session", such as in group chats or public channels.

In the main session, Clawdbot has full system permissions. It can read and write files, execute bash commands, control the browser, and operate the calendar. Since only you are having a conversation with it at this time, the risk is controllable.

But in the non - main session, Clawdbot will automatically switch to the "sandbox mode", where it will execute all operations within a Docker container.

It's like preparing a "training ground" for it. In this isolated environment, it can still execute commands but cannot access the real data on your computer. Even if someone tries to attack your system through malicious commands in a group chat, Clawdbot will only execute them in the sandbox and will not affect the real environment.

Another technical challenge is the coordination between the Gateway and the Tools.

The gateway is the control center running on your device. It is responsible for receiving messages from various chat apps, maintaining session states, scheduling the AI model, and managing scheduled tasks. The tools are the various capabilities that Clawdbot can call.

The cooperation between the two is not simple. When you send a message to Clawdbot in a chat app, the gateway needs to do the following things:

1. Identify which session the message comes from (whether it's the main session or a group chat)

2. Decide what level of permissions to use for processing (full permissions or sandbox mode)

3. Send the message to the AI model to understand the intention

4. The AI model returns the tools and parameters to be called

5. The gateway then executes these tools

6. Finally, return the result to you.

Throughout the process, the gateway needs to handle multiple sessions simultaneously. Each session may have different permission configurations, and it also needs to ensure that the order of messages is maintained and the state is not lost.

This is like a traffic police officer who needs to coordinate vehicles coming from different directions, ensuring efficiency while avoiding collisions.

To solve this problem, Clawdbot has designed a "session model". Each session has its own state, history records, and permission configurations.

Although AI inference still needs to call the APIs of Anthropic or OpenAI, the routing logic and automated control are completely local, ensuring privacy and controllability. This architectural design allows Clawdbot to leverage the powerful capabilities of cloud - based large models while maintaining the privacy principle of local priority.

The memory of the model is limited, but Clawdbot is a product for long - term continuous work. Therefore, it stores all interaction records in a local folder in Markdown format, forming a persistent memory similar to Obsidian's knowledge base.

It will remember your preferences, habits, and ongoing projects. Each conversation can continue the context, rather than starting from scratch like traditional AIs.

At the same time, Clawdbot also has a certain "self - upgrading" ability. For example, when you tell it "I want you to be able to generate images", it will research on its own how to access the image - generation API, improve its own code, and then tell you "I can now generate images".

Clawdbot's official team calls this process "Self - Expanding Skills".

It will first search the official skill library and the skills already installed locally by the user. Only when there are no available resources will it conduct a full - scale API research and code writing.

After the skill upgrade and expansion are completed, Clawdbot will go through three steps: syntax checking, permission verification, and function trial - run. If the test fails, it will automatically roll back and report the problem to the user without affecting the existing functions.

In addition, Clawdbot is an active AI assistant. Through cron scheduled tasks, it can actively send messages to you. It can automatically send a schedule summary every morning, summarize the weekly work progress every Friday afternoon, monitor specific conditions, and send timely reminders.

This initiative breaks the traditional passive - response mode of AIs. You no longer need to remember to ask it; instead, it will actively reach out to you at the right time.

Running Clawdbot doesn't require expensive hardware. Although there are many stories online about people buying Mac minis to run Clawdbot, in fact, for basic scenarios such as chatting, summarizing, and API calls, a VPS that costs $5 per month is sufficient.

Only when you need to run a local large model or perform heavy - duty automated tasks do you need to consider local hardware like the Mac mini. The rush to buy Mac minis is more of a manifestation of social proof and geek culture rather than a necessity.

Not only can it operate a computer, but from a technical implementation perspective, Clawdbot supports more than 50 integrations, including Gmail, GitHub, Obsidian, Notion, Todoist, Spotify, Philips Hue smart lights, etc.

That means you can control all the smart furniture in your house with just one text message.

Clawdbot can connect almost all the digital tools in your life and work, becoming a real digital - life hub. And since it is open - source, the community is constantly contributing new skills and plugins.

02

Why is Clawdbot so popular?

Clawdbot's popularity stems from the fact that it solves the core pain point of existing AI tools: excessive friction.

This problem may sound simple, but it is the biggest obstacle preventing AI from truly integrating into daily life.

Traditional AIs require you to open a web page, switch tabs, and copy and paste results. Each step interrupts your workflow.

Suppose you are writing an email and suddenly need AI to polish a paragraph. You have to switch to the ChatGPT tab, paste the content, wait for the response, and then copy it back.

This process may only take a few seconds, but it disrupts your thinking and makes the whole process discontinuous. In contrast, Clawdbot is right in your chat app. You send a message and get a response without any context switching.

Another major reason is privacy.

In today's era when AI services generally rely on the cloud, Clawdbot's concepts of local priority, open - source transparency, and user controllability hit the core demands of the technical community. Your conversation records, personal preferences, and automated scripts are only stored on your own device and can be viewed and modified at any time.

Having "data sovereignty" in your own hands is a huge attraction for users who value privacy.

A deeper reason is that Clawdbot represents a new software paradigm. Traditional software is static, where developers write the functions and users passively use them. In contrast, Clawdbot is dynamic and malleable.

You can let it learn new skills at any time and make it transform itself according to your needs.

This is actually a new model of "Software as a Chat (SaaC)", where each user can have a personalized assistant based on their unique language style and needs.

From the community's reaction, Clawdbot's popularity also reflects a collective longing. People have been waiting for Siri to become smarter for more than a decade and for Alexa to become truly useful for nearly a decade, but the products of these big companies have remained at the level of "setting an alarm or checking the weather".

Clawdbot, through an open - source approach and the power of the community, has achieved in just a few weeks what these big companies promised but failed to deliver for years.

This atmosphere of "everyone is a product manager and everyone is a developer" has resonated strongly in the technical community.

It's true that there is nothing extremely profound about Clawdbot technically. In essence, it is a "harness" that converts the powerful capabilities of the model into practical assistant functions. Without the progress of these models, Clawdbot would not be possible.

However, its product concept hits people's pain points.

The key to an AI assistant is not the AI but the assistant. So it shouldn't be a separate website; it should be an integral part of your daily tools.

More importantly, Clawdbot has expandable capabilities. Some people let Clawdbot summarize the health data of their Whoop bracelets every morning, some let it automatically handle insurance - claim emails, and others let it monitor server status and automatically fix problems when they occur.

These are not official preset functions but are created by users according to their own needs.

03

Yuanbao's opportunity has arrived

Clawdbot's success proves that an AI assistant should be integrated into the existing communication tools of users rather than requiring users to install a new app.

In China, this tool is WeChat.

WeChat is not only China's largest communication tool; it is also an integral part of the modern - life workflow.

People chat, work, make payments, and shop on WeChat. Almost all aspects of digital life can happen on WeChat.

If Yuanbao can exist in the form of a contact in the address book, it will have a natural advantage in reaching users. Users don't need to download a new app or learn a new interface. They just need to send a message to Yuanbao as if they were sending a message to a friend. This zero - threshold way of use is unparalleled by any independent app.

Moreover, WeChat's advantages go beyond that.

WeChat has a payment and mini - program ecosystem, which means Yuanbao can do more than Clawdbot.

Clawdbot can help you look up information, write code, and manage your schedule, but it's difficult for it to complete closed - loop operations such as payments, bookings, and shopping.

If Yuanbao can integrate with WeChat Pay and mini - programs, it can truly become an all - purpose assistant. You can ask Yuanbao to order takeout, buy movie tickets, pay utility bills, and make doctor appointments. All these operations can be completed in the conversation without switching to other apps.

Clawdbot stores its memory as Markdown files, which is already quite advanced. But Yuanbao can build a higher - dimensional and more accurate personal knowledge base within the WeChat ecosystem using chat records, pictures, emojis, subscribed official accounts, and WeChat - subscribed services.

WeChat records your social relationships, consumption habits, and interest preferences. If Yuanbao can make reasonable use of this data (of course, with user authorization and privacy protection), it will understand you far better than any other AI assistant.

This ability of persistent memory will turn Yuanbao from a temporary tool into a long - term partner. The longer users use it, the better it will understand them, and the more value it can bring.

Clawdbot's initiative to care about users is also an inspiration for Yuanbao. Through the information bound to WeChat, Yuanbao can do even more.

Pushing personalized schedules and weather information every morning is a basic function. But Yuanbao can also monitor express delivery logistics and actively remind users to sign for packages, regularly summarize consumption bills and provide financial advice, and recommend content or services based on user habits.

For example, if Yuanbao notices that you have been frequently searching for a certain topic recently, it can actively push relevant official - account articles or video - account content.

This is not just a fantasy. From a technical implementation perspective, it is completely feasible for Yuanbao to replicate Clawdbot's core capabilities.

Clawdbot's architecture is not complex. As mentioned earlier, a message gateway, a large - language model interface, and a tool - calling system are all it needs.

Tencent has all these technologies and may even do a better job. Moreover, Tencent has its own Hunyuan large model. If Tencent is willing to invest resources, Yuanbao can definitely become the Chinese version of Clawdbot, or even do better.

However, for Yuanbao to succeed, it needs to solve several key problems.

The first is privacy and trust. Clawdbot is popular largely because it is open - source and prioritizes local storage, allowing users to have full control over their data.

As a Tencent product, Yuanbao is bound to be a cloud - based service. How to make users believe that their data is safe and will not be misused is a huge challenge.

The second is definitely openness. Clawdb