How Vector works
Vector is an agent: it reads your design, plans the change, makes it through the application’s own interface, and checks the result.
The working loop
- 1
Read the design
Vector starts from the open document, its features, and your selection, plus anything you attach to the message.
- 2
Plan the change
Vector breaks the request into steps. When a request is ambiguous, such as which of two faces a reference should use, it asks instead of guessing.
- 3
Act through the application
It runs scripts through the application’s API: Fusion’s MCP server, the SOLIDWORKS add-in, or the Bambu Studio adapter. When a dialog needs a click, it can use the application’s interface directly.
- 4
Check the result
It takes screenshots of the application window to confirm what changed, and can measure the geometry it built.
- 5
Report every step
Each action appears in the conversation, and the result is in your model as native features.
Native results
When Vector models, it creates sketches, features, joints, and parameters the same way you would, so the timeline or feature tree stays editable and you can change anything by hand afterward.