Dollhouse Bridge
Bridge is the integration runtime for connecting a Dollhouse-powered session to external communication systems.
It is designed so people can keep working in the tools they already use while the AI system runs with explicit safety and approval boundaries.
Current Scope
- Connector-based architecture for chat platforms
- Session dispatcher for routing requests to active runtime sessions
- File-based inbox/outbox contract for predictable message flow
- Approval-aware execution paths aligned with Dollhouse safety controls
Architecture At A Glance
Bridge separates platform APIs from session logic:
- Platform connector reads and normalizes messages
- Connector writes to a local inbox queue
- Dispatcher routes work to the session backend
- Session produces output and writes to outbox
- Connector posts back to the external channel
That boundary makes connectors easier to add and test without rewriting core runtime behavior.
Current and Planned Connectors
- Active architecture and implementation work around Zulip-based flows
- iMessage integration paths in progress
- Planned expansion to additional channels (including Slack, Discord, Matrix)
Why Teams Care
Bridge lets teams run AI-assisted workflows where team communication already happens.
That means:
- less tab-switching during operations
- clearer review loops for sensitive actions
- easier handoff between humans and agent workflows
How It Fits The Portfolio
Bridge is the channel and workflow layer that complements:
- DollhouseMCP for element orchestration
- MCP-AQL for compact operations and policy-aware endpoint boundaries
- Collection for reusable behavior building blocks