Subagents on Cursor
January 2026
Subagents on Cursor are gonna be game changers
The insane speed benefits that the cursor has, combined with semantic search methods, sets it apart from Claude Code
Yes, Claude Code excels in a variety of tasks, i.e., large codebase refactoring, introducing a scaffolding for a project, etc
However, the semantic search used in Cursor to gather high-fidelity context really stands apart when solving bugs quickly
In Claude Code, you have to use explore sub-agents to speed up the process, and you need to do it every time to build up relevant and good context. You start a new chat to solve bugs of a similar nature (maybe we can use skills here), and you need to rebuild your context. Claude Code is good when the context is properly built up, including mentioning things that it might have missed in its exploration.
However, the ability for Cursor to pick up schemas or relevant files without explicit mention is extremely beneficial, especially since fast sub-agents like Composer models can be invoked by good orchestrator/planner agents like Opus 4.5, which is mind-blowing!
Really excited to be trying it out :)