Cursor Opens Side Chats, Letting Side Questions Run Separately During Long AI Tasks

What’s new

Cursor introduced Side Chats and Conversation Search in its official changelog on July 10. The version is listed as 3.11, and the update includes Side Chats, Conversation Search, redesigned project and repo pickers, and Cloud Agent hooks.

The main point of this news is that AI coding tools are moving beyond simply generating answers and toward managing long agent tasks. As a single task gets longer, questions like “Can I ask something else for a moment?”, “What conclusion did we reach before?”, and “Where is this agent running?” become real productivity issues.

What problem do Side Chats solve?

Side Chats open a new auxiliary conversation next to the main agent chat. According to Cursor, users can create one with /side, /btw, or the plus button at the top of the chat panel, and the new side chat brings in context from the main conversation.

In simple terms, it is like asking another question at the desk next to the agent that is handling a large task. For example, users can check questions like “Is this option right?”, “What alternative libraries are available?”, or “Would changing direction now be risky?” without interrupting the main task. It is a way to keep the main task running while moving only exploration, checks, and branch questions to the side.

Why Conversation Search matters

Cursor says users can search agent transcripts in the Agents Window through the command palette, meaning Cmd+K. The flow is not just about finding conversation titles or PR numbers; it also makes the contents of past agent chats searchable.

Inside an existing conversation, users can also search with Cmd+F and move through results in a long transcript while viewing a match counter. Cursor says its local search index scales quickly to thousands of conversations. It is a sign that conversation history in AI coding is becoming reusable working knowledge, not a one-time log.

Pickers and Cloud Agent hooks also changed

This changelog also includes redesigned project and repo pickers. Cursor says users can create projects and continue connecting GitHub, GitLab, and Azure DevOps inside the picker, while search scope is divided by work location, such as This Computer, Cloud, or a specific remote machine.

Cloud Agent hooks were also expanded. In addition to existing hooks around tool execution or file and shell work, Cursor added hooks that observe or control the agent conversation itself, such as prompts, responses, thinking, subagents, compaction, and turn completion. Hook names such as beforeSubmitPrompt, afterAgentResponse, subagentStart were also disclosed. This points toward giving teams better ways to monitor agent results and, when needed, build their own correction loops.

What users should watch carefully

Still, more features do not mean every team should immediately use them in the same way. If Side Chats multiply, the reasoning behind decisions can become scattered. As Conversation Search becomes more convenient, teams also need to think carefully about how much sensitive work history they leave behind. Cloud Agent hooks are powerful, but without team policy they can make notification and review flows too heavy.

In short, this Cursor update is closer to an AI coding management feature. From now on, developers need to design not only around model performance, but also around conversation branching, history search, execution location, and hook policy. The longer we hand work over to AI, the more important tools become for managing that work without interrupting it.