Physical AI News (Japan)

Physical AI, robotics and compute — reported from Japan.

日本語
· Product launches· 日本語で読む

Claude Platform Lists API Features, Explains Browser and Computer Use, Data Controls

Japanese-language documentation for Claude Platform lists features available through its API. It describes browser and desktop use, reusable Agent Skills and asynchronous batch processing. It also covers geographic controls for inference and stored data.

What happened

Japanese-language documentation on platform.claude.com lists features that work across the Claude Platform API. It describes browser and desktop use, Agent Skills, batch processing and data residency. The page’s publication date could not be confirmed.

Why it matters

For developers, the documentation offers a basis for considering how to connect Claude’s responses to a user-provided browser or desktop environment, or to custom code. Data residency is described as a feature for controlling where inference takes place and where stored data is kept, making it relevant when processing and storage need to be managed separately.

Analysis

The page appears to organize features and connection points for using Claude through an API rather than announce new features. Some features connect model responses to external processes through tool calls. With custom tools, Claude requests a call, code runs it and returns the result. Agent Skills are described as reusable packages containing instructions and other material, presenting task-specific procedures and the code that performs actual operations as separate components. For physical AI, it is important not to misread the scope of the computer-use features. The computer use described in the documentation operates a desktop environment using screenshots, a mouse and a keyboard. Browser use is also a mechanism for connecting to a user-side environment. These features handle on-screen interactions and are distinguished in the documentation from functions that control robot joints or sensors. Therefore, this feature list alone does not provide grounds to infer Anthropic’s plans for robotics development or control of physical robots. OpenAI disbands independent robotics team, moves researchers to applied divisions, covered by this newspaper, was a report about organizational structure. The material here, by contrast, explains API features and is a different kind of information from developments in corporate robotics teams or physical robot development. Putting the two side by side does not provide evidence of a connection between Anthropic and robotics development. The feature structure points to a design in which a developer-provided execution environment, tools used by the model and code that returns results are connected. Batch-processing pricing terms and the ability to separately control the locations of inference and stored data also concern how the API is operated and data is managed. However, the documentation does not provide details in this list about the performance of individual features, the full terms of use or actual deployment cases. To discuss a connection with physical AI, it would be necessary to separately verify whether there are control methods for physical systems, sensor inputs, safety designs and operational track records. See the original documentation for details.

Sources

Translated from the Japanese original: /news/anthropic-claude-platform-no-api-kinou-wo-bunsho-de-shoukai-8ba014. The Japanese version is authoritative if they differ.