Cursor CLIにデバッグモードと/btwコマンドのサポートを追加 CLI Debug Mode and /btw Support

Cursor has rolled out an update to its command-line interface that introduces a dedicated debug mode and adds support for a new /btw command. While Cursor is best known as an AI-assisted code editor, the company has steadily expanded its CLI capabilities in recent months, allowing agents to be invoked directly from the terminal. The latest changes continue that trajectory.

The debug mode appears designed to surface verbose logs and expose the internal flow of an agent's execution when it is run from the command line. Because modern AI coding agents chain together reasoning steps, tool invocations, and file system operations, pinpointing exactly where a run went off the rails has become a practical necessity. A built-in debug switch should make it easier for developers to inspect prompts, tool calls, and intermediate state without having to instrument the process externally — a meaningful quality-of-life improvement for anyone integrating the CLI into larger workflows.

The second addition, support for /btw, is less self-explanatory. The name likely stands for "by the way," suggesting a mechanism for injecting supplementary context into an ongoing conversation or for bridging context between the editor and the CLI. Cursor has previously emphasized parity between its IDE and terminal experiences, and a growing vocabulary of slash commands fits that pattern. The exact behavior of /btw was not detailed in depth in the changelog entry, but it appears to extend the conversational surface area available to users working outside the editor.

Taken together, the two changes reinforce a broader theme in Cursor's recent development: treating the CLI not as a secondary surface but as a first-class environment for agent-driven coding. Debugging tools and richer command syntax are the kinds of features that matter most when an agent is running unattended or as part of an automated pipeline, rather than under the watchful eye of a developer in an IDE.

The competitive backdrop is hard to ignore. Anthropic's Claude Code, Google's Gemini CLI, and OpenAI's Codex CLI have all staked out positions as terminal-native AI coding agents, and each has been iterating quickly on ergonomics, tool use, and scripting integration. For Cursor, whose origins are firmly in the editor, hardening the CLI is a way to remain relevant in scenarios where an IDE is impractical or unavailable — continuous integration jobs, remote servers, container-based development environments, and ad-hoc automation tasks among them.

There is also a strategic logic to making the CLI more debuggable. As teams begin to embed AI agents in build pipelines and code review automation, failures need to be reproducible and inspectable. A black-box agent that occasionally produces incorrect patches is difficult to operationalize; one whose execution trace can be examined after the fact is far easier to trust. Debug mode, in that sense, is less about developer convenience in the moment and more about the kind of observability that production deployments demand.

The /btw command, if it does indeed allow lightweight context injection, would complement that direction by giving users a way to nudge agents mid-run without restarting a session or rewriting a prompt. Such affordances have become common across competing tools, and standardizing them in Cursor's CLI would help close any remaining feature gaps with rivals.

Neither change is dramatic on its own, and the changelog entry is brief. But the cumulative effect of incremental updates like this one is what determines whether a CLI graduates from a convenient wrapper into a tool teams actually rely on. With debug mode and /btw, Cursor appears to be steadily filling in the practical details required for that transition, even as the headline competition continues to focus on model quality and agent autonomy.