Agentic Systemsintermediate12 min · 9 steps
Real-World Agents: AutoGPT to Agentic RAG
Six well-known agent systems, each a different recipe from the same ingredients: autonomous loops, task queues, coding agents, graphs, role-playing crews and retrieval that reasons.
tool callresponsedatarequestsignal
Key ideaStep 1 of 9
Same ingredients, six different recipes
Every agent mixes the same parts: a model that reasons, tools, memory and a loop. Well-known real-world agents differ in which part they put in charge: the model's own prompts, a task queue, a test suite, a graph, a team of roles, or the retrieval step.
Studying them side by side is the quickest way to learn the design space, including the failure modes that pushed later systems toward more structure.
| System | What's in charge | Demonstrates |
|---|---|---|
| AutoGPT | The model's self-prompting loop | Autonomous goal pursuit |
| BabyAGI | A task list it rewrites | Task creation and prioritisation |
| OpenDevin | Code, tests and a sandbox | Autonomous software engineering |
| LangGraph | A state graph you define | Stateful, conditional workflows |
| CrewAI | A crew of role-playing agents | Multi-agent collaboration |
| Agentic RAG | Reasoning about information needs | Retrieval as a decision |
Lesson 15 of 15 in Agentic Systems