Feature request for product/service
Cursor IDE
Describe the request
Current AI coding workflows still put too much responsibility on the developer to determine what context the agent needs.
In a large codebase, knowing which files to attach is often harder than asking the question itself. Even with search, grep and dependency traversal, the agent can still start from incomplete context and spend tokens discovering or correcting its assumptions.
It would be useful to have a repository-level context layer that can:
• identify files relevant to a task
• resolve cross-module dependencies
• understand call relationships
• surface architectural context
• provide the agent with evidence for why those files are relevant
The goal isn’t simply “more context”, but the right context with less manual selection and less exploratory token waste.
This seems particularly important as codebases become larger and agents take on more complex debugging and architectural tasks.