文章摘要
Anthropic has launched significant new tool-use capabilities for its Claude AI assistant, introducing a secure Sandbox and an MCP Tunnel feature.
- The Sandbox provides Claude with a secure execution environment to run Python code and perform web browsing, enhancing research and problem-solving.
- The MCP Tunnel is based on the Model Context Protocol, enabling developers to integrate Claude with custom internal tools, databases, and APIs.
- These integrations allow Claude to securely interact with external systems, expanding its functionality beyond its core training data.
- The new features are designed to empower Claude to act as a more capable and autonomous agent for complex, real-world tasks, leveraging external resources.