Anthropic has introduced four new beta capabilities on its API platform designed to enable developers to build more powerful AI agents: code execution tool, MCP connector, Files API, and extended prompt caching up to one hour. These features work alongside Claude Opus 4 and Sonnet 4 models to eliminate custom infrastructure requirements for enterprise AI development.
The code execution tool transforms Claude from a code-writing assistant into a data analyst capable of running Python code in sandboxed environments for computational results and data visualisations. Organisations receive 50 free hours daily, with additional usage priced at $0.05 per hour per container. The tool enables end-to-end analytical tasks including financial modelling, scientific computing, business intelligence, document processing, and statistical analysis within single API interactions.
The MCP connector eliminates the need for custom client code when connecting Claude to remote Model Context Protocol servers. The API automatically handles connection management, tool discovery, error handling, and authentication while accessing third-party tools from providers including Zapier and Asana. This reduces complexity for building tool-enabled agents by managing the entire integration process automatically.
The Files API streamlines document management by allowing developers to upload files once and reference them across multiple conversations, particularly beneficial for applications requiring large document sets like knowledge bases or technical documentation. The API integrates with the code execution tool, enabling Claude to process uploaded files and generate outputs like charts during execution.
Extended prompt caching now offers one-hour time-to-live options compared to the standard five-minute caching, providing 12x improvement for long-running agent workflows. This enhancement reduces costs by up to 90% and latency by up to 85% for long prompts, making previously cost-prohibitive long-running agent applications economically viable at scale.
These capabilities enable comprehensive AI agent development without custom infrastructure investment. Enterprise applications can maintain context over extended periods for multi-step workflows, complex document analysis, and system coordination. The features join existing tools like web search and citations to create a complete enterprise AI development toolkit, supporting scalable agent deployment across business functions.