Introducing Dollhouse Research
Dollhouse Research is the home for products, protocols, and applied AI systems work from DollhouseMCP Inc.
The goal is practical: publish composable AI tools people can run, modify, and combine in real workflows.
What Is Here Today
- DollhouseMCP as the core composition platform
- Dollhouse Collection as the community catalog for reusable elements
- MCP-AQL as the protocol surface for token-efficient, structured operations
- Bridge as the cross-channel runtime for remote workflows
- AILIS as architecture research for shared AI systems language
- Merview as a supporting tool for high-clarity markdown and diagram review
The Building Blocks We Use
Dollhouse work is organized around four terms:
| Term | Meaning |
|---|---|
| Elements | Reusable units (persona, skill, template, agent, memory, ensemble) |
| Agents | Execution-focused elements with lifecycle and approval-aware runtime flows |
| Ensembles | Coordinated element packs for repeatable multi-part behavior |
| Stacks | End-to-end solution compositions built from multiple elements and ensembles |
Element files -> Session activation -> Ensemble coordination -> Agent execution -> Stack outcome -> Publish and reuse
Elements -> reusable files
Agents -> runtime execution
Ensembles -> coordinated bundles
Stacks -> full end-to-end workflows
A Concrete Stack Example
Elemental Surveys (currently early/private development) is a strong early example of the stack idea:
- research personas for methodology and synthesis
- survey and analysis skills
- report and scorecard templates
- memory elements for benchmarks and research references
- orchestration agents and ensembles for full survey workflows
The core idea is simple: build with readable pieces, then combine them into complete capabilities.
Why This Approach Works
Dollhouse projects favor transparent composition over black-box prompting.
That means:
- you can inspect and edit the pieces directly
- you can swap components without rewriting everything
- you can share, review, and version behavior as files
- you can grow from one element to an ensemble to a full stack over time
What Changed Recently
The current 2.0.0 refactor path for DollhouseMCP improves reliability and launch readiness:
- cleaner service boundaries
- stronger logging, metrics, and diagnostics
- safer execution and approval paths
- better performance and caching behavior
MCP-AQL continues to reduce registration overhead by consolidating operations into semantic endpoint paths with runtime introspection.
Where To Start
- Open Projects for the full product map.
- Start with DollhouseMCP.com for installation and platform docs.
- Use MCPAQL.com for protocol context.
- Browse the Dollhouse Collection for reusable components.