We’re excited to introduce Subagents! Subagents are independent agents specialized to handle discrete parts of a parent agent’s task.
How it works
Subagents run in parallel, use their own context, and can be configured with custom prompts, tool access, and models. The result is faster overall execution, more context in your main conversation, and specialized expertise for each subtask.
Cursor includes default subagents for researching your codebase and executing parallel work streams. These will automatically start improving the quality of your agent conversations
You can also define custom subagents. Learn more in our docs.
Useful for
Parallel execution: break a complex plan into parallel work streams
Deep research: spin up multiple subagents to explore different parts of your codebase
We’d love your feedback!
How are subagents handling your complex tasks?
Have you tried creating custom subagents yet?
What specialized subagent configurations would be useful for your workflow?
If you’ve found a bug, please post it in Bug Reports instead, so we can track and address it properly, but also feel free to drop a link to it in this thread for visibility.
Amazing! Is there a way to override/modify the default subagents? I see in the docs project-defined overrides user-defined, but I cannot find the default subagents anywhere, they are not listed in cursor settings either.
For the moment, no. We will be closely listening for feedback on this (so thank you for sharing!)
The hope is that using models like Composer for subagents (like “Explore”) will significantly speed up tasks that don’t require high-latency models that, despite excelling in certain areas, are slower to respond. This has proved out really well in our testing!
The hope is that using models like Composer for subagents (like “Explore”) will significantly speed up tasks that don’t require high-latency models that, despite excelling in certain areas, are slower to respond. This has proved out really well in our testing!
That would be cool if we can switch explore agent to use Gemini 3 Flash instead of overpriced Composer 1
Subagents do not receive User Rules. So for each Subagent you write down its name, description for the calling Agent, model and its personal rules of conduct.
Can you elaborate on the documentation. If you want to use a claude sub-agent, can you use both .cursor and .claude? I’m used to putting everything in my .cursor folder, but now I see you have .claude and .codex locations. Will the sub-agent not pick to use a claude model if it’s defined in .cursor folder?
