Elements, Ensembles, and Stacks
How composable Dollhouse elements map to real application behavior and operational workflows.
Research
Research at Dollhouse is practical by design. We publish work that helps people model AI systems clearly, compose capabilities safely, and move from experiments to repeatable production behavior.
MCP-AQL
Operation-surface simplification research for lower token overhead and cleaner client integration.
Current Themes
- Human-readable AI composition and control
- Agent lifecycle safety, review, and authorization
- Protocol ergonomics and integration cost reduction
- Spec-driven interfaces that stay in sync with live implementations