1 min lesson
Choose when a subagent helps
Choose when a subagent earns a separate context, then decide whether it should block or run in the background.
Step 1 of 2
A subagent is worth the startup cost when its separate context keeps a large, independent task out of the parent conversation.
A subagent starts with a clean context and returns its final result to the parent. Use one for long exploration, independent verification or work that can run in parallel. Cursor has built-in Explore, Bash and browser subagents for codebase searches, command output and browser automation through MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition.. A foreground subagent blocks until it returns. A background subagent lets the parent continue and writes its output under ~/.cursor/subagents/. Project subagents live in .cursor/agents/*.md and can set description, model, readonly and is_background in frontmatter. Subagents share the parent checkout unless you ask for isolated project copies, so parallel writers can overwrite each other. A short single-purpose action usually belongs in a Skill instead.
- Entry point
- Agent delegation or .cursor/agents/*.md
- Source
- Cursor Subagents docs
Check clean context, run modes, custom fields and checkout isolation in Cursor's Subagents docs.