The headline addition is side chats that inherit context from the main conversation but stay durable and separate. Cursor also added transcript search, in-chat find, better project and repo pickers, and new hooks such as beforeSubmitPrompt, afterAgentResponse, and subagent lifecycle controls.
This is a meaningful quality-of-life upgrade for anyone running agents on real codebases where context sprawl becomes the bottleneck. Searchable transcripts and programmable hooks make it easier to audit agent behavior, recover prior work, and build tighter automation around long sessions.
Use side chats for tangents like dependency research, architecture alternatives, or API checks while your main coding thread stays clean. If you run cloud agents for team workflows, wire the new hooks into observability or policy layers so you can monitor prompts, responses, and subagent behavior.
Read Original Post →