--- title: Agent Architectures subtitle: Explore all available agent architectures and compare their capabilities slug: guides/agents/architectures no-image-zoom: true hide-toc: true layout: overview --- MemGPT agents Agents that can edit their own memory Sleep-time agents Memory editing via subconscious agents Low-latency (voice) agents Agents optimized for low-latency settings ReAct agents Tool-calling agents without memory Workflows LLMs executing sequential tool calls Stateful workflows Workflows that can adapt over time ## Comparing the architectures **Unsure of which architecture to use?** Consider starting with our default agent architecture (MemGPT), which is highly autonomous and has long-term self-editing memory. You can constrain the behavior to be more deterministic (ie more "workflow-like") by adding [tool rules](/guides/agents/tool-rules) to your agent. | Architecture | Reasoning Traces | Tool Calling | Tool Rules | Persistent Messages | Long-term Memory | Usecase | |--------------|------------------|--------------|------------|---------------------|------------------|---------| | [MemGPT agents](/guides/agents/architectures/memgpt) | ✓ | ✓ | ✓ | ✓ | ✓ | Long-running (perpetual) stateful agents | | [Sleep-time agents](/guides/agents/architectures/sleeptime) | ✓ | ✓ | ✓ | ✓ | ✓ | Async (subconscious) memory processing | | [Low-latency (voice) agents](/guides/agents/architectures/low-latency) | ✓ | ✓ | ✓ | ✓ | ✓ | Stateful agents with latency constraints | | [ReAct agents](/guides/agents/architectures/react) | ✓ | ✓ | ✓ | ✓ | - | Simple memory-less tool-calling agents | | [Workflows](/guides/agents/architectures/workflows) | ✓ | ✓ | ✓ | - | - | Predefined, sequential processes | | [Stateful workflows](/guides/agents/architectures/stateful-workflows) | ✓ | ✓ | ✓ | - | ✓ | Workflows that can adapt over time |