ChatGPT has undergone a major upgrade to its container-based code execution environment, now supporting a wide range of new capabilities that significantly expand its utility for development and data analysis tasks. As of January 26, 2026, the feature—previously known as the Code Interpreter and later rebranded as Advanced Data Analysis—has evolved into a much more powerful and flexible toolset, now referred to informally as ChatGPT Containers. One of the most notable changes is the ability to run Bash commands directly within the sandboxed container. Previously, only Python code execution was supported, with shell commands limited to being invoked through Python’s subprocess module. Now, users can issue native Bash commands, enabling full control over the environment, file system operations, and process management. In addition to Bash, ChatGPT can now execute code in multiple programming languages including Node.js (JavaScript), Ruby, Perl, PHP, Go, Java, Swift, Kotlin, C, and C++. While Rust and Docker are not yet available, the presence of registry references for these languages in the container’s environment suggests they may be coming soon. A key advancement is the ability to install packages using pip and npm. Despite the container’s lack of direct outbound internet access, these package managers function through a secure internal proxy system hosted at packages.applied-caas-gateway1.internal.api.openai.org. Environment variables like PIP_INDEX_URL, UV_INDEX_URL, and NPM_CONFIG_REGISTRY are configured to route package installations through this proxy, allowing safe and reliable dependency management. Another major addition is the container.download tool, which enables ChatGPT to fetch files from public URLs and save them directly into the container’s filesystem. This allows for seamless workflows where the AI can locate data online—such as Excel files, CSVs, or logs—download them, and then process them using Python, Bash, or other languages. The tool is designed with security in mind; it only permits downloads from URLs that were either explicitly provided by the user or surfaced through web searches, preventing prompt injection attacks from exfiltrating sensitive data. The container environment also includes a range of other tools, such as python.exec for internal reasoning, web.run for internet access, file_search.msearch for querying connected data sources, and container.exec for running arbitrary commands. The full toolset is accessible via a structured interface, with each tool having a defined signature and purpose. Despite these powerful new features, OpenAI has yet to publish official documentation or release notes detailing the changes. This lack of transparency makes it difficult for users to understand the full scope of capabilities, limitations, and security safeguards. The absence of clear guidance also raises questions about how the system handles privacy, data retention, and potential edge cases. Nonetheless, the upgrade marks a significant leap forward in AI agent capabilities. By combining language support, package installation, file downloading, and shell access in a single sandboxed environment, ChatGPT now functions more like a full-featured development environment. This enables users to write, test, and debug code across multiple languages, automate data workflows, and interact with external resources—all within a single conversation. The feature is available even on free accounts, suggesting broad rollout. As the ecosystem evolves, users will likely see further enhancements, including support for additional languages, better error handling, and more robust documentation. For now, the new capabilities represent a major step toward truly autonomous AI-powered development.