YOYO Claw Can Really Code This Time: HONOR Laptops Add Coding and Video Agents

HONOR announced its laptop upgrades for this month on September 14. Two changes are most noteworthy for developers: YOYO Claw now has professional coding capabilities and has added a video creation agent. In addition, HONOR has introduced a floating ball, previewing and sharing of generated content, third-party service access, and a range of other features focused on everyday use.

This is not a simple chatbot revamp. HONOR wants to move YOYO Claw from an AI entry point that users must actively open toward becoming a work assistant built into the computer.

Of course, it is too early to equate it with mature coding agents. The update announces capability directions and product features, but HONOR has not disclosed which specific models are being used, whether its coding capabilities support complete codebase understanding, whether it can autonomously run tests, or how different tasks are divided between the cloud and the device. For developers, this looks more like a product-form completion than a model-capability release.

From Chatting to Participating in Development

Previously, YOYO Claw's product narrative focused mainly on on-device agents, built-in skills, and ease of use. When HONOR introduced the related technology in April this year, it emphasized features such as out-of-the-box availability, scenario-based intelligent recommendations, and local security protection, and associated it with the so-called lobster trend. Simply put, users do not need to study complicated deployment processes, API keys, or plugin configurations themselves; they can access built-in AI capabilities as soon as they turn on the computer.

The addition of professional coding capabilities in the September update means that YOYO Claw is beginning to enter the areas where developers actually work every day: reading code, writing code, fixing bugs, generating scripts, explaining errors, adding tests, and even collaborating continuously around a project.

The key issue here is not simply whether the model can generate code, but whether YOYO Claw can place coding tasks in a sustainable context. Development work differs from ordinary Q&A because code is not a one-time answer. A seemingly correct function may be unusable because of project dependencies, directory structure, the runtime environment, or historical design decisions. A genuinely useful coding agent needs to understand project boundaries, know which files it can modify, know how to verify the results of its changes, and continue converging after errors occur.

HONOR has not yet released further technical details about its coding capabilities, so it would be unwise to overinterpret this as a complete Agentic Coding product. From a user-experience perspective, however, the capability has at least three practical uses:

1. Small scripts and automation tasks: For example, batch-renaming files, cleaning spreadsheets, converting image formats, calling local tools, or generating a script to process logs.
2. Learning and troubleshooting: Users can ask YOYO Claw to explain errors, break down function logic, or compare two implementation approaches, reducing the need to switch back and forth between search engines and documentation.
3. Prototype development: For product managers, designers, and independent developers, generating a web prototype, data-processing tool, or small internal application first can significantly shorten the distance from an idea to a runnable version.

What these tasks have in common is that they do not demand the same depth of context as a large production project, but they are highly sensitive to interaction efficiency. A tool that can be called quickly on a computer, lets users see generated results, and allows them to continue making changes is often more convenient than simply copying a question into a web chat box.

The Ceiling of Coding Capabilities Is Not Determined by the Model Alone

The most important aspect of this update to watch may not be whether it can generate several hundred lines of code, but how it plans to handle computer permissions.

When AI moves from answering questions to modifying files, running commands, and installing dependencies, the risks rise at the same time. A coding agent might accidentally delete files or execute dangerous commands; the dependency packages it generates may pose supply-chain risks; and when reading a local project, it could bring secrets, configuration files, and user data into the model context.

When HONOR previously promoted YOYO Claw, it mentioned device-level security protection and an independent security lobster, emphasizing monitoring of high-risk operations and requiring secondary confirmation for sensitive operations such as logins and payments. If this approach can be fully extended to coding scenarios, it will be far more valuable than simply providing a more attractive chat interface.

Ideally, a coding agent should have at least the following layers of control:

• Permission levels: Project read-only access, code modification, command execution, and network access should be authorized separately rather than granting all permissions at once.
• Operation previews: When AI is preparing to modify files or execute commands, it should first display the changes and their scope of impact.
• Rollback: Every change should be reversible, preferably in conjunction with version control or system snapshots.
• Sensitive-information isolation: Environment variables, SSH keys, cloud-service credentials, and similar content should not be read or uploaded by default.
• Execution-result feedback: The agent should not merely say that the code has been written; it should also show whether tests passed and which steps failed.

This is also the dividing line between YOYO Claw and pure chat products. Developers are not afraid of AI making mistakes; they are afraid that it will make mistakes without leaving a trace or providing a way to recover.

Video Creation Agent: Expanding from Writing Code to Creating Content

In addition to coding capabilities, HONOR has added a video creation agent. The official description says that it provides professional and entertaining AI creation capabilities, but no complete feature list has been made public. What can be confirmed at present is that the product entry point has appeared; the specific scope of support for video generation, editing, voice-over, subtitles, and asset processing remains subject to the actual version being rolled out.

Placing this update alongside the coding capabilities reveals YOYO Claw's product strategy: rather than serving only one specialized profession, it packages different tasks as callable scenario-based agents.

For ordinary users, the video creation agent might be used for short-video scripts, storyboard planning, subtitle organization, rewriting materials, and simple finished-video production. For developers and content teams, it is more likely to become part of an automated pipeline—for example, converting product update notes into a video script, generating voice-over copy according to a fixed template, and then handing it off to an editing tool.

However, video agents are more prone than text-based Q&A to creating a gap between expectations and results. When a user asks for a technology-themed product promotional video, actual implementation also involves cinematography, pacing, brand assets, subtitle standards, music licensing, and output aspect ratios. AI's ability to generate something that looks like a video does not mean it can consistently produce a video ready for publication.

Therefore, the key to whether YOYO Claw's video capabilities are useful is whether they support editable, reusable intermediate results rather than merely giving users a final file. For example, can the script, shot list, asset list, subtitles, and voice-over be adjusted separately? Can users lock the brand colors, character designs, and visual style? Can one video quickly be adapted into landscape, portrait, and versions for different platforms? These details determine whether it is a toy or a production tool.

The Floating Ball and Preview Sharing Complete the Last Mile

Among this update's additions, the floating ball looks the least like a core capability, but it may have the greatest effect on usage frequency. HONOR has added a floating ball for YOYO Claw that provides shortcuts to frequently used functions.

AI products often get stuck in an awkward position: users know they can do many things, but each use requires opening the app, finding the entry point, and describing the request again. The floating ball places the entry point on the system desktop, lowering the cost of invocation. For developers, copying an error message, selecting a piece of code, or asking AI to explain the current page could ideally all require fewer steps.

The newly added preview and sharing functions are equally important. In the past, if AI-generated content remained in a chat window, users still had to copy, download, or organize it manually. YOYO Claw's ability to quickly view and share generated content indicates that HONOR is beginning to treat outputs as files, images, videos, or other deliverables.

This may sound like basic functionality, but whether an agent product can ultimately fit into a workflow often depends on these basic infrastructures. If generated results cannot be previewed smoothly, users cannot determine whether they are usable; if they cannot be shared, collaboration gets stuck inside a personal window; without a history log, users also find it difficult to know where their points and invocation costs have gone.

This update also adds user-account details, allowing users to view their history of point consumption. At the same time, it introduces an activity that gives points for inviting friends, with both the inviter and the invitee receiving points. The former addresses usage transparency, while the latter clearly carries a growth-marketing purpose. The addition of a dedicated SVIP user-operations channel also indicates that YOYO Claw is establishing a user-service system independent of system features.

Third-Party Access Keeps the Agent from Only Looping Internally

YOYO Claw has added third-party integrations with Sunlogin and Tongcheng Tongxin. The reference materials do not disclose the specific invocation methods or permission scopes of these integrations, but in terms of product direction, HONOR is attempting to connect YOYO Claw to more external services.

If an agent can only generate text inside its own chat box, its capability ceiling is quickly reached. An agent that genuinely saves time needs to read information, call tools, and deliver results to platforms that users are already using. Remote control, information collaboration, content publishing, and device operations could all become areas for further expansion.

However, third-party integrations also bring more complex security issues. Each additional service means another set of accounts, permissions, and data flows. If HONOR wants developers and enterprise users to adopt these capabilities, it will need to publish clearer explanations of permissions, data-handling methods, log-retention policies, and mechanisms for revoking abnormal operations.

This is particularly important because local execution does not automatically mean absolute security. Even if files remain on the computer, as long as the agent can log in to third-party services on the user's behalf, read page content, or perform remote operations, the risk has already expanded from the local environment to the account system.

Magic View Is Changing Too: The Desktop Is Becoming an AI Canvas

Magic View is also being upgraded alongside YOYO Claw. HONOR has added AI-generated wallpapers and AI-generated cute-pet features, allowing users to customize their desktop backgrounds and entertaining desktop pets.

These features are clearly more consumer-oriented. They will not directly improve code compilation speed, but they reflect another way HONOR understands AI entry points: AI should not only be hidden inside a tool; it can also exist continuously within the desktop environment. Wallpapers and cute pets are low-risk, highly visible ways of showcasing AI capabilities, making them suitable for helping users form usage habits.

From the perspective of industry competition, smartphone and PC manufacturers are both competing for system-level AI entry points. Comparing a single generation result is of limited significance; the real competition is whether AI can cover work, creation, entertainment, and device control at the same time, while allowing users to avoid repeatedly switching between applications. By placing coding, video, and desktop content in the same round of updates, YOYO Claw is carrying out precisely this kind of horizontal expansion.

Which Models Can Be Upgraded?

This upgrade covers a relatively broad range of devices, including the following HONOR laptop models:

• HONOR WIN gaming laptops H9 and H7
• HONOR MagicBook Pro series 2026
• HONOR MagicBook Digital series 2026
• HONOR Laptop X Plus series 2026, 2025, and 2024
• HONOR MagicBook Art 14 2025 and 2024
• HONOR MagicBook Pro 16 2025 and 2024
• HONOR MagicBook Pro 14 2025
• HONOR MagicBook 14 series 2023
• HONOR Laptop X series 2026 and 2025
• HONOR Laptop X series Ryzen editions 2026 and 2024
• HONOR MagicBook X14 Ryzen edition 2023
• HONOR MagicBook X14 2023
• HONOR MagicBook X16 Ryzen edition 2023
• HONOR MagicBook X16 2023

The specific rollout schedule, system-version requirements, and feature differences between models remain subject to HONOR's subsequent announcements and the updates actually received by each device. Although hardware coverage is important, it cannot simply be understood to mean that all models will receive exactly the same on-device capabilities. The local computing power, memory, and NPU configurations of different processors may affect response speed and available functions.

Assessment: More Like a System-Level Entry-Point Upgrade Than a Model Leap

Taken together, the significance of YOYO Claw's September update lies in completing the agent's usage loop: professional tasks now have a coding entry point, content tasks have a video entry point, everyday invocation has a floating ball, outputs have preview and sharing, usage costs have point details, and external capabilities continue to expand through third-party integrations.

Its strengths are clear: it is closer to the computer, has a lower barrier to entry, and is suited to handling scripts, content, and desktop tasks. If the on-device security, permission controls, and device-level protection that HONOR previously emphasized can be implemented in actual operations, YOYO Claw may indeed be better suited to continuous work on a computer than a simple web-based chat tool.

Its weaknesses are equally clear: the publicly available information is not yet sufficient to prove that its coding capabilities can replace professional development tools, nor is it enough to determine whether the video agent can meet stable production standards. For developers working in complex engineering environments, IDE plugins, code-repository context, terminal execution, test verification, and version control remain indispensable.

In other words, YOYO Claw is not trying to redefine AI models this time; it is competing for the first screen AI reaches after entering the computer. Whether it can stay there will ultimately depend on three things: whether its generated results are reliable enough, whether its permission boundaries are clear enough, and whether it connects smoothly enough with the tools users already have.

If HONOR can subsequently open up more complete developer interfaces, skill-extension mechanisms, and model-selection capabilities, YOYO Claw may have a chance to grow from a manufacturer-preinstalled feature into a genuine PC agent platform. For developers, the most important question right now is not whether it can write a piece of code, but whether it can connect code, files, tools, and collaboration workflows in a secure and controllable way.

References

1. ITHome: HONOR Announces September Upgrades for Its Laptops — Summarizes this month's YOYO Claw and Magic View upgrade items and supported models.
2. Official HONOR YOYO Claw Introduction — Introduces the PC version of YOYO Claw and information about HONOR's related agent products.
3. Zhihu: HONOR Debuts the Preconfigured Lobster AI Agent Technology YOYO Claw — Provides additional information about YOYO Claw's built-in primary lobster, sub-lobsters, and scenario coverage.