Feature request for product/service
Cursor Web
Describe the request
Hi Cursor team 
I’d like to request subagents support for Cloud Agents, including both standard Cloud Agents and Long-Running Cloud Agents.
Why this matters
For larger workflows, a single agent quickly becomes a bottleneck: it has to juggle planning, coding, debugging, testing, and docs all at once. Subagents would let a “manager” agent delegate work in parallel or in a structured pipeline, making Cloud Agents much more effective for real-world projects.
What I’m asking for
- Subagents for Cloud Agents
Ability for a Cloud Agent to spawn and coordinate multiple subagents.
Each subagent can have its own role/instructions (e.g., “test writer”, “refactorer”, “docs”, “dependency auditor”).
- Subagents for Long-Running Cloud Agents
Same capability, but with long-running context and state.
Subagents can be created/terminated dynamically as tasks evolve.
Key use cases
Parallelization: one subagent writes tests while another implements the feature, and a third reviews for style/security.
Separation of concerns: dedicated subagents for “architecture decisions”, “bug triage”, “performance”, “documentation”.
Controlled autonomy: manager agent keeps the global plan; subagents operate with narrower permissions/scopes.
Additional request: Launch Long-Running Cloud Agents from the Cursor IDE (like Cloud Agents)
Launching Cloud Agents from the Cursor IDE already works well. I’d like the same ability for Long-Running Cloud Agents—launch them directly from the IDE UI/command palette, without a separate flow.
Important detail: Slash commands are backed by local agent files (.cursor/agents)
When I say “launch with subagents,” I don’t mean a literal /subagents command — that was just a placeholder.
In practice, many of us define subagents as agent files (e.g. Markdown) inside .cursor/agents, and then invoke them via slash commands like /my-agent-name (where the name comes from the agent file / definition).
Request: Make these local agent definitions usable by Cloud Agents (IDE + Web options)
For Cloud Agents (and Long-Running Cloud Agents), it would be great to support one of these approaches:
Option A — Bundle local agent files from the IDE when creating a Cloud Agent
On Cloud Agent creation in the IDE, automatically include (or allow selecting) agent definitions from .cursor/agents.
This would make /my-agent-name subagents available immediately in the Cloud Agent session.
Option B — Allow uploading/managing agent files in Cursor (web)
Let users upload and manage their agent definitions (the same files used locally in .cursor/agents) in the Cursor web UI.
Cloud Agents could then access the same curated set of subagents without relying on IDE-side bundling.
Either option would make subagents much more practical for Cloud Agents, especially for teams that already maintain a library of internal agent definitions.
Thanks! Happy to share a concrete folder example (.cursor/agents), expected behavior, or help test in beta.