Multi-Agent System
Also known as: MAS, multi-agent architecture, agentic AI, agent network
A network of specialized, autonomous AI agents that collaborate and divide complex tasks among themselves to solve problems too difficult for a single model.
Source: Common AI/ML terminology; Wikipedia
Primary reference ↗A Multi-Agent System (MAS) coordinates multiple specialized AI models — each with distinct roles, tools, and areas of expertise — to complete tasks that require sequential reasoning, parallel execution, or domain-specific knowledge that no single model handles well alone.
Architecture Pattern
A typical three-layer MAS architecture used in biological research:
User Query
↓
Planning Agent ← Understands intent, decomposes query, routes to specialists
↓
Execution Subagents ← Domain experts: tissue biomarker agent, antibody agent, etc.
↓
Synthesis Agent ← Aggregates results, resolves conflicts, formats response
↓
Validated ResponseWhy MAS for Biological Data
Single-model (“chatbot”) approaches struggle with biological research because:
- Queries require multi-step reasoning across heterogeneous databases
- Different query types (tissue biomarkers vs. antibody search) need different tool sets
- Validation requires cross-referencing multiple data sources simultaneously
- Context windows fill up quickly with large database payloads
Specialized agents can each focus on a narrow, well-defined task — improving accuracy and reducing hallucinations compared to one model handling everything.
Key Challenges
- Orchestration: Getting agents to correctly hand off work without losing context
- Observability: Tracing which agent made which decision and why
- Cost: Each agent call incurs LLM API costs; poorly orchestrated systems multiply these
- Consistency: Agents must apply the same validation standards (e.g., tau thresholds) independently