InertiaRSS Track and read blogs, news, and tech you care about
Read Original Open in InertiaRSS

Recommended Feeds

博客园 - 司徒正美
V
V2EX
T
Tailwind CSS Blog
有赞技术团队
有赞技术团队
aimingoo的专栏
aimingoo的专栏
Apple Machine Learning Research
Apple Machine Learning Research
IT之家
IT之家
Blog — PlanetScale
Blog — PlanetScale
A
About on SuperTechFans
月光博客
月光博客
T
The Blog of Author Tim Ferriss
宝玉的分享
宝玉的分享
Martin Fowler
Martin Fowler
博客园 - 聂微东
The GitHub Blog
The GitHub Blog
V
Visual Studio Blog
WordPress大学
WordPress大学
酷 壳 – CoolShell
酷 壳 – CoolShell
Engineering at Meta
Engineering at Meta
GbyAI
GbyAI

阮一峰的网络日志

科技爱好者周刊(第 396 期):互联网通信的替代方案 科技爱好者周刊(第 396 期):互联网通信的替代方案 - 阮一峰的网络日志 科技爱好者周刊(第 395 期):软件开发的第三种方式 科技爱好者周刊(第 395 期):软件开发的第三种方式 - 阮一峰的网络日志 科技爱好者周刊(第 393 期):脑腐状态 科技爱好者周刊(第 392 期):axios 投毒与好莱坞式骗术 科技爱好者周刊(第 391 期):AI 的贫富分化 科技爱好者周刊(第 390 期):没有语料,大模型就是智障 套壳中国大模型撑起500亿美元估值?扒一扒 Cursor 的"套壳"疑云 科技爱好者周刊(第 389 期):未来如何招聘程序员 科技爱好者周刊(第 388 期):测试是新的护城河 科技爱好者周刊(第 387 期):你是领先的 科技爱好者周刊(第 386 期):当外卖员接入 AI 字节全家桶 Seed 2.0 + TRAE 玩转 Skill 科技爱好者周刊(第 385 期):马斯克害怕中国车企吗? 智谱旗舰 GLM-5 实测:对比 Opus 4.6 和 GPT-5.3-Codex 科技爱好者周刊(第 384 期):为什么软件股下跌 科技爱好者周刊(第 383 期):你是第几级 AI 编程 Kimi 的一体化,Manus 的分层 科技爱好者周刊(第 382 期):独立软件的黄昏 AI native Workspace 也许是智能体的下一阶段 科技爱好者周刊(第 381 期):中国 AI 大模型领导者在想什么 科技爱好者周刊(第 380 期):为什么人们拥抱"不对称收益" 科技爱好者周刊(第 379 期):《硅谷钢铁侠》摘录 我如何用 AI 处理历史遗留代码:MiniMax M2.1 升级体验 科技爱好者周刊(第 378 期):预测是新的互联网热点 科技爱好者周刊(第 377 期):14万美元的贫困线 科技爱好者周刊(第 376 期):太空数据中心的争议 科技爱好者周刊(第 375 期):一扇门的 Bug 终于有人做了 Subagent,TRAE 国内版 SOLO 模式来了 科技爱好者周刊(第 374 期):6GHz 的问题 VS Code 使用国产大模型 MiniMax M2 教程 科技爱好者周刊(第 373 期):数据模型是新产品的核心 国产大模型接入 Claude Code 教程:以 Doubao-Seed-Code 为例 科技爱好者周刊(第 372 期):软件界面如何设计 大模型比拼:MiniMax M2 vs GLM 4.6 vs Claude Sonnet 4.5 科技爱好者周刊(第 371 期):一个乐观主义者的专访 科技爱好者周刊(第 370 期):正确的代码高亮 错误处理:异常好于状态码 科技爱好者周刊(第 369 期):Tim 与罗永浩的对谈 科技爱好者周刊(第 368 期):不要这样管理软件团队 一天之内,智谱和 Anthropic 都发了最强编程模型 科技爱好者周刊(第 367 期):Nano Banana 的几个妙用 科技爱好者周刊(第 366 期):旧金山疯狂的 AI 广告 科技爱好者周刊(第 365 期):流量变现正在崩塌 科技爱好者周刊(第 364 期):最难还原的魔方 科技爱好者周刊(第 363 期):最好懂的神经网络解释 科技爱好者周刊(第 362 期):GitHub 工程师谈系统设计 科技爱好者周刊(第 361 期):暗网 Tor 安全吗? 科技爱好者周刊(第 360 期):Dan Wang 的新书
Zero-install "Cloud Shrimp Farming": ArkClaw User Guide
阮一峰 · 2026-03-12 · via 阮一峰的网络日志

I. Introduction

Have you been bombarded with "lobster" (OpenClaw) these past two days?

Its news is everywhere, even the NPC delegates and Xinhua News Agency are talking about it. It's truly astonishing that an AI software can cause such a big reaction.

People are enthusiastic, the free offline installation event is overcrowded, while the "paid installation" business online is booming.

Many people may still not know that there is now the simplest way to use lobsters:ArkClaw.

So simple that you don't need to worry about installation.Because this is a no-installation solutionIt comes directly with shrimp, ready to use.

I also started using it yesterday and can't wait to share my initial impressions with everyone. For those who haven't used it before, you can also consider it as a tutorial for "Zero Barrier to Lobster," to see what lobster is all about.

2. What is ArkClaw?

Here's the situation, old readers might remember, I was before the Spring FestivalEvaluatedByte's newly released Seed 2.0 model.

I said in the article that this is ByteDance's strongest base model currently, and the phone DouBao uses it, the test performance is very good.

Later, my colleagues from Byte gave meCoding PlanPackage, for easy continued testing of this model, various AI programming tools can call its API (of course, the package also includes other domestic models, which are also freely usable).

On Monday this week, I suddenly found that ByteDance's Coding Plan package had included a bundled service, which is ArkClaw.

I asked customer service and learned thatif you sign up for the Coding Plan now, you can use Lobster for free..

In other words, as long as you use ByteDance's AI programming package, you don't need to spend an extra penny—ByteDance provides a remote host with Lobster installed, which you can use freely.

It should be noted that the Coding Plan is divided into two packages: lite (¥9.9 for the first month) and Pro (¥49.9 for the first month). The lite package can only be experienced for free for 7 days, only the Pro package can be used long-term with ArkClaw.

III. Cloud Lobster Raising

ArkClaw belongs to the category of "cloud-raised shrimp" (also known as "cloud crayfish"), which involves installing a crayfish (OpenClaw) on a cloud server within the Volcano Ark (ByteDance's AI cloud service brand). The "ark" in its name refers to "ark."

Besides "cloud-raised shrimp," crayfish can also be installed on local computers.

Those unfamiliar with it might wonder about the differences between the two, so I'll briefly explain.

First, you need to know that OpenClaw is an automation software , its function is to allow users to describe their needs in natural language, it finds methods that meet the needs through large models and then completes them automatically.

When it is installed on a local computer (your laptop), it makes it convenient to automate operations on local files and devices, such as "find photos taken today last year" or "turn off the living room smart lights and check the power consumption in the past week."

When installed in the cloud, it can interact with various network services 7x24 hours, such as "automatically generating a 30-character content summary when receiving an email and sending a notification to the phone."

So, if you need to automate network service operations and require long-term online presence or daily scheduled runs, then it's suitable to use "cloud shrimp farming."

Four, ArkClaw Basic Operations

4.1 Interface

Let me show you what ArkClaw looks like.

Enter Console and click "Create Now" to create a shrimp instance.

After creation, it's already installed and ready to use.

The interface is very simple, just a dialog box. ArkClaw has customized the official control console for shrimp, simplifying the operation interface.

4.2 Capture Information

You can have a conversation with the AI model in the dialog box, which is no different from the usage of other models.

For example, we can make it capture information.

As you can see, since it captures dynamic content, the model came up with many implementation plans and successfully completed it in the end.

Remember, ArkClaw is a remote host. Any server technology solution that can be used, it can use, and it is more powerful than a Lobster installed on a regular personal computer.

4.3 Send Messages

After obtaining the information, the Lobster can send this information to a mobile phone.

Currently, ArkClaw supports binding with WeCom, DingTalk, and Lark. Among them, Lark, being a self-developed product, has the simplest binding process, which is fast and convenient—scanning a QR code is all it takes. The other two require more complex operations, seeofficial documentation

Click the "Lark Pairing" button above the dialog box. (The "Message Channel" button before it is used to bind WeCom and DingTalk.)

The system will open a terminal window, output a QR code. After scanning with Lark, a robot can be created to bind with the current ArkClaw instance.

Through this robot, you can chat with the current ArkClaw instance on your phone.

You can also send messages to your phone from your computer via the ArkClaw web console.

After entering the above command on the computer, a push notification will be sent to your phone (see the image below).

4.4 Scheduled Tasks

We can also specify the time and frequency for the lobster to perform certain tasks, known as scheduled tasks.

First, use natural language to set up scheduled tasks in the dialog box.

After setup is complete, your phone will receive messages daily.

To delete a scheduled task, you also use natural language to issue commands.

V. Skill and Other Settings

5.1 Skill

The capabilities of the Lobster itself are limited, and it will always encounter some problems it doesn't know how to handle. At this point, its abilities can be extended through Skill (Skills), which greatly increases its utility.

What is Skill? Simply put, it's a package containing instructions and examples, used to teach the model how to complete specific tasks.

There are already many Skills written and shared by others online. Just pick the ones you need and have the Lobster load them to expand corresponding capabilities. Website ClawHub.ai collects Lobster-specific Skills, with nearly 20,000 available.

I originally wanted to use Xiaohongshu SKill as an example to demonstrate how lobsters learn to write Xiaohongshu. But the official account posted something yesterday.AnnouncementIt's been too many people doing this recently, and now accounts are being banned.

Then let's take another example.

The screenshot above shows using natural language to have the lobster download and install from the ClawHub website.The skills of Gaode Map (amap).

The lobster didn't know how to use Gaode Map at first, but after acquiring this skill, it learned and could query information from it. The specifics and details of this skill can beCheck its homepage.

When using it, you can directly describe your needs in natural language, and the model will automatically load and call the required skills.

The screenshot above is from Amap, checking real-time traffic conditions.

5.2 Other Settings

Other features of ArkClaw are in the "Settings" menu (below), such as adjusting the underlying model.

Any model provided by the Coding Plan package can be used here.

The "Settings" menu also has two very useful features.

One is "Open Terminal," which opens a terminal window on the web, allowing you to directly operate the remote host where ArkClaw is located via the command line.

As can be seen from the terminal window screenshot above, ArkClaw is based on the Ubuntu system.

The other is "Configure Network Storage". In certain situations, you may need to upload/download files to ArkClaw. This feature allows the current host to bind with the object storage service TOS of火山引擎, which is equivalent to having an unlimited-capacity network storage.

VI. Summary

This is the main content of my first use of ArkClaw yesterday.

My feeling is thatit indeed greatly simplifies the use of crawlerswithout installation, ready to use out of the box, making the operation of crawlers simple and intuitive. Loading and calling skills through natural language is also very natural and smooth.

Its biggest strength lies in its deep integration with the ByteDance ecosystemworking together seamlessly: underlying Seed 2.0 model + Lark push + Huoshan Engine cloud disk, completely eliminating the need for complex configuration.

It is a service bundled with ByteDance Coding Plan, no additional payment required. Compared to building a "cloud crawler" from scratch, the cost of cloud hosting and AI models is saved, which is a significant amount of money.

As a programmer, this AI programming Coding Plan + Cloud Lobster ArkClaw hosting bundle is still quite attractive.

(End)