Ability to create a plan and launch multiple sub-agents

Feature request for product/service

Cursor IDE

Describe the request

Problem

In a monorepo / enterprise product, even “small” features usually touch multiple layers (frontend, backend, APIs, DB, microservices). Today, planning + implementation tends to be serialized because a single agent is trying to reason across many folders and contexts at once. That slows delivery and increases context-switching mistakes.

Proposed feature

  • Add a Monorepo Execution Plan workflow that can:
  • Ingest a monorepo structure (or selected folders).
  • Generate a top-level execution plan for a feature request.
  • Split the plan into folder-scoped tasks (frontend, backend, serviceA, serviceB, etc.) based on:
    • auto-detected architecture, and/or
    • user direction (“split by /apps/web, /services/auth, /services/billing”).
    • Spawn parallel sub-agents (or “workers”) where each sub-agent is:
      - restricted to a folder scope
      - equipped with shared context (global requirements, API contracts, schema changes)
      - able to run tests/builds for its scope (if supported)

Reconcile outputs with a coordinator agent that:

  • validates cross-layer contracts (routes/types/schemas)
  • resolves conflicts (e.g., API shape mismatch)
  • creates a final integrated PR set (or staged PRs per folder)

Why it matters (value)

  • Enables true concurrency for cross-layer features.
  • Reduces errors by scoping each agent’s context to the code it owns.

Provides repeatable “enterprise-grade” delivery: plan → tasks → parallel execution → contract validation.

1 Like