Research

Research areas

Persistent multi-agent systems

How do you build AI agents that maintain context across session boundaries? Our sleep/wake model with SQLite-based memory is one approach. We're researching:

• Optimal memory schemas for long-running agents
• Context compression and relevance scoring
• Cross-session learning and reflection patterns

Agent-to-agent communication

How should autonomous agents collaborate? Our A2A protocol handles direct messaging, broadcasts, escalations, and priority overrides. Research questions:

• Optimal communication patterns for different organizational structures
• Conflict resolution between agents with competing priorities
• Information propagation speed vs. noise reduction

Organizational learning

How does an AI organization accumulate institutional knowledge? Our knowledge library model lets individual agent experiences benefit every agent. Open questions:

• Knowledge quality scoring and deprecation
• Cross-domain knowledge transfer
• Avoiding knowledge drift and hallucination propagation

Autonomous operations

How much autonomy should AI agents have? Our authority-level model defines decision scopes with automatic escalation. Research includes:

• Optimal authority boundaries for different operational contexts
• Human-in-the-loop integration patterns
• Trust calibration over time

Applied research

All our research is grounded in real deployments:

Publications

Coming soon. We're preparing our first publications on:

• Multi-agent architectures for industrial operations
• Persistent memory systems for LLM-based agents
• Open source AI platforms for the maritime industry