Compare Architectures
Side-by-side: use case, requirements, latency, cost, strengths, weaknesses, failure modes, example — and when to choose each.
Workflow vs AgentSingle agent vs Multi-agentRAG vs Fine-tuningVector search vs BM25Function calling vs MCPDirect API integration vs MCPPlan-then-execute vs ReActShort-term memory vs Long-term memoryLLM-as-judge vs Deterministic evaluatorsHuman-in-the-loop vs Human-on-the-loopSupervisor vs PipelinePrompt caching vs Semantic cachingRouter vs SupervisorFramework vs No framework
| Supervisor | Pipeline | |
|---|---|---|
| Use case | Dynamic delegation: which specialist runs next depends on results | Fixed sequence of stages, each transforming the previous output |
| Requirements | Coordinator agent, worker agents, shared state, hand-off schema | Ordered stages with typed inputs / outputs |
| Latency | Coordinator call before and after every worker | Sum of stages; stages can stream |
| Cost | Coordinator re-reads context each turn | Each stage sees only what it needs |
| Strengths | Flexible routing, retries, and re-delegation | Simple, testable stage by stage, predictable |
| Weaknesses | Supervisor becomes the bottleneck and a single point of failure | Cannot revisit earlier stages without explicit loops |
| Failure modes | Coordinator loops between workers; lost task state | Error in stage 1 silently degrades every later stage |
| Example | Research task delegating to search, code, and writing agents | Transcribe → extract entities → summarize → publish |
| Choose this when | Choose a supervisor when the sequence of specialists is not knowable in advance. | Choose a pipeline when the stages are fixed; it is a workflow with LLM steps and should be your default. |