Cursor 3.11 ships side chats, agent transcript search, and five new cloud hooks

Cursor 3.11, released July 10, 2026, is easy to misread as a UI refresh: a way to open a second chat, a search box for old conversations, tidier pickers. The more useful reading is ergonomic. Since April, Cursor has been pushing developers toward running many agents at once, and 3.11 is the release that finally addresses what that workflow breaks: your ability to ask a side question without derailing a run, and to find what any past agent actually did.

The release follows an aggressive shipping cadence. Cursor 3.9 (June 29) brought the iOS app for remote agent management. 3.10 (June 30) added Team MCPs and organization groups. Now 3.11 rounds out the cycle with tooling for the supervision debt that parallel-agent work creates.

Side chats and transcript search

Side chats are created with /side, /btw, or the plus button. Each one is a durable, full agent conversation, not a throwaway popup. You can revisit it later and @-mention it to pull its context back into the main thread. For developers running long agent sessions, this means no more losing context when a colleague asks a question.

Agent transcripts are now searchable at scale. A local search index, exposed via Command+K in the Agents Window, searches actual message content. Cursor says the index scales to thousands of conversations. Within a single open transcript, Command+F now searches text with a match counter, solving the problem of scrolling through hours of agent output looking for a single decision point.

New cloud hook events

Five new hook events for cloud agents expand observability: beforeSubmitPrompt, afterAgentResponse, afterAgentThought, stop, and subagentStart. These let teams observe agent reasoning and output on cloud VMs and build self-correcting loops around them. With this release, 14 of the 20 documented hook events now run on cloud agents.

The project and repo pickers were also rebuilt around location: search is scoped by This Computer, Cloud, or Remote Machine rather than one undifferentiated global list. That only makes sense as a design priority if a meaningful share of users now hold projects across all three.