Cursor integration
Cursor integration is a Sphere feature, not part of the open SCAN notation or toolkit. It connects a component on the architecture board to a coding harness that can implement it.
Early access. This feature is not self-serve yet. Contact us to request a demo.
What it does
Sphere stays the control plane; Cursor is an execution worker. Selecting a component and approving a plan does not hand Cursor the whole repository — it hands Cursor one bounded, reviewable task.
- Select a component on the board. Sphere reads its SCAN description, contracts, and connected repository.
- Sphere drafts an implementation plan — scope, contract impact, and a set of small work items — grounded in that architecture context and the repository's existing code.
- You review and approve the plan. Nothing is written until you do.
- Sphere runs Cursor CLI in an isolated branch/worktree against one approved work item at a time, streams its progress, and enforces the allowed file scope.
- Sphere independently verifies the result (build, lint, tests, contract compatibility) before offering to create a commit or pull request.
Cursor doesn't decide product scope and can't silently change the architecture. If the implementation implies an architecture change, Sphere raises it back as a proposed SCAN update for review — it isn't applied automatically.
Why this exists
SCAN describes what the system is. Sphere plans, governs, and verifies the work. Cursor performs the bounded code changes. Keeping that boundary means the architecture diagram stays trustworthy even as an AI harness is writing code against it.
Get access
Reach out at spherescan.io@gmail.com to request a demo.