OpenAI has just added a powerful new capability to its “deep research” feature in ChatGPT: a GitHub connector that allows users to analyze codebases directly from the platform.

Announced Thursday, this marks the first official “connector” for ChatGPT’s deep research tool, which was originally designed to comb the web and other sources to compile detailed research reports. With the GitHub integration (currently in beta), developers can now query code repositories and engineering documentation directly—bringing AI deeper into the software development workflow.

The feature will roll out to ChatGPT Plus, Pro, and Team users in the coming days, with Enterprise and Edu support on the way, OpenAI said.

This GitHub integration comes amid a wider industry trend: AI companies are increasingly linking chatbots to external platforms to make them more context-aware and practical. Anthropic, for instance, recently launched “Integrations” for its Claude chatbot, enabling similar platform connections.

OpenAI had previously offered plugin functionality for ChatGPT, but has since shifted toward “custom GPTs”—tailored chatbot experiences. The GitHub connector builds on that evolution.

“I often hear that users find ChatGPT’s deep research agent so valuable that they want it to connect to internal sources, not just the web,” noted Nate Gonzalez, OpenAI’s Head of Business Products, in a LinkedIn post. “That’s why we’re introducing our first connector.”

With the GitHub connector, users can:

Ask questions about a codebase or technical documentation
Break down product specifications into actionable tasks
Summarize code structure and identify patterns
Learn how to implement APIs with real-world code examples

As with any AI feature, accuracy isn’t guaranteed—ChatGPT can still hallucinate or generate incorrect information. OpenAI stresses that the tool is intended as a productivity booster, not a substitute for human expertise.

Importantly, ChatGPT will only access GitHub repositories that users have permission to view. OpenAI confirmed that organizational settings and permissions will be respected, so only explicitly shared codebases will be visible to the assistant.

The update is part of a broader push by OpenAI to enhance its developer tools. Recent moves include launching Codex CLI, an open-source terminal-based coding assistant, and improving the ChatGPT desktop app’s ability to read code within popular developer environments.

Programming continues to be a major focus for OpenAI. Reports suggest the company has agreed to acquire Windsurf, an AI-powered code assistant startup, for $3 billion.

Also on Thursday, OpenAI rolled out new fine-tuning options for developers. Organizations can now customize the o4-mini “reasoning” model through reinforcement fine-tuning, a method that uses task-specific grading to improve accuracy. Fine-tuning is also now available for the GPT-4.1 nano model.

While o4-mini fine-tuning is limited to verified organizations, GPT-4.1 nano customization is open to all paying users. OpenAI began requiring verification for access to certain models and features in April, citing the need to prevent misuse and ensure responsible deployment.

Source