Cursor 2.0 - Our first coding model and new interface for agents
by•
Two big updates that make Cursor the best place to work with agents: our first coding model, Composer, and a new interface for working with many agents in parallel.
Also included in this release: browser for Agent, voice mode, improved code review, and more!
Replies
Thanks to your team, coding has become so much easier for me.
Lately though, I feel AI getting a bit dumb , maybe I’m just not using it right?
I'm a newbie and I just saw that Cursor released Composor1. Could you please tell me in which scenarios Composor1 is suitable to use, and in which scenarios Sonet4.5 is more appropriate?
Minara
Crashed many times if I open 4-5 cursor and run claude code inside after several hours. Have no Idea what happend. 32G RAM
Subagents running parallel with their own context is the part of Cursor 2.0 that changed my workflow the most. One agent handles the refactor, another writes the tests, and they don't pollute each other's windows. Visual Editor closing the gap between design tweaks and production code is a nice bonus, but the agent autonomy slider is what keeps me from context-switching back to a terminal.
Taskade AI Mobile: Supercharge Workflows
background agents and subagents running in parallel is exactly where this needs to go. the single-threaded "ask one question, wait for response" pattern doesnt match how real development works.
the contextual understanding of the full codebase is still Cursors biggest moat. other tools generate code well but Cursor actually understands the relationships between files. that matters a lot for refactoring and debugging.
pairing this with Claude Code for bigger agentic tasks has been our best dev workflow so far. Cursor for exploration and inline edits, Claude Code for multi-file changes.