customers define agents as fully autonomous systems that operate independently over extended periods, using various tools to accomplish complex tasks. Others use the term to describe more prescriptive implementations that follow predefined workflows. At Anthropic, we categorize all these variations as agentic systems, but draw an important architectural distinction between workflows and agents: Workflows are systems where LLMs and tools are orchestrated through predefined code paths. ワークフローとは、LLMとツールがあらかじめ定められたコードの流れ に従って連携するシステムです。 Agents, on the other hand, are systems where LLMs dynamically direct their own processes and tool usage, maintaining control over how they accomplish tasks. 一方で、エージェントは、LLMが自らプロセスやツールの使い方を動的 に判断・制御しながらタスクを遂行するシステムです。つまり、どうやっ て目的を達成するかをLLM自身が柔軟に決める点が特徴です。 - Building effective agents, Anthropic
Amazon Bedrock Agents Ease of use & balance of capabilities OpenAI Agents SDK Enterprise-level functionalities Amazon Bedrock Agents • Some agent frameworks come with built-in support for IAM configuration and integration with knowledge bases. • While setting up IAM and similar components may require more effort than using prepackaged frameworks, it brings you closer to enterprise-grade deployment. Support for (Pre-defined) Workflows High Floor / High Ceiling Memory, Human feedback, Code Interpreter, Integration, Fault tolerance • Defining clear workflows can help ensure greater system stability. • (Increasing the agent's level of autonomy (Agency Level) isn't always better — even OpenAI and Anthropic have acknowledged this.) * Multi-agent communication / Streaming is becoming available in all frameworks? The higher the level of abstraction in an agent, the harder it becomes to maintain fine-grained control. Frameworks that are easy to use with minimal code often come with this downside. 1 2 3 X