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
| Single agent | Multi-agent | |
|---|---|---|
| Use case | One task domain, a manageable tool set, one context | Distinct specializations, separate contexts, or parallel sub-tasks |
| Requirements | One loop, one prompt, tools | Topology (supervisor / pipeline / swarm), message schema, shared state |
| Latency | One sequential chain of calls | Coordination overhead; can be lower when sub-tasks run in parallel |
| Cost | Lower: no coordination tokens | Higher: each hand-off re-reads context; supervisors add calls |
| Strengths | Simple to trace, evaluate, and secure | Context isolation, specialization, parallelism |
| Weaknesses | Context bloat when too many tools or domains | Hand-off errors, duplicated work, supervisor bottleneck |
| Failure modes | Wrong tool among 40; overflowing context | Agents disagree, loop between each other, lose task state |
| Example | Support bot with 8 tools and a knowledge base | Research pipeline: search agent → synthesis agent → critic |
| Choose this when | Choose a single agent first; add tools and better prompts before adding agents. | Choose multi-agent when a single context provably cannot hold the task, or sub-tasks are independent enough to parallelize. |