GitHub Copilot Chat on the Web: Tool Call Visibility, Chat Export, and Better Repo Attachment

GitHub Copilot Chat on the Web Gains Transparency and Export Features

On February 4, 2026, GitHub shipped a set of improvements to Copilot Chat at github.com, targeting three areas where the web-based chat experience had room to grow: transparency into what Copilot is doing, portability of chat history, and usability of repository context attachment.

Real-Time Tool Call Visibility

When Copilot receives a message and begins processing it, the agent may invoke multiple tools internally — running searches, reading files, calling APIs — before formulating a response. Previously, this intermediate activity was invisible to the user. The update surfaces these tool calls in the chat interface in real time, showing clear details about each step as it happens.

This transparency serves two practical purposes. First, it helps developers understand the reasoning and steps behind a response — useful when debugging unexpected output or evaluating whether Copilot found the right context. Second, it allows users to course-correct mid-session if they can see the agent is heading in the wrong direction, rather than waiting for a completed response to realize the approach was off. The tool call details also include references generated by each tool call, making it easier to trace exactly what sources informed an answer.

Chat Export in JSON and Markdown

Copilot Chat conversations can now be downloaded for external reference. To export a conversation, users open the "Manage chat" menu and select the Export option, then choose between JSON and Markdown format. The JSON format is suited for programmatic use or integration with other tools, while Markdown provides a clean, human-readable record that can be pasted into documentation, shared with teammates, or archived for future reference.

Improved Repository Recommendations

When attaching a repository to a Copilot Chat session — giving the model additional codebase context — GitHub now surfaces better repository recommendations during the search process. The improved recommendations reduce the friction of locating the correct repository, particularly in organizations with large numbers of repos or similar naming conventions.