Curated example · Editable connections

AI agent architecture
you can explore.

See how an orchestrator, model, tools, and memory work together.

Curated example

Follow the connections.

Select a component to explore

User Request

Defines the task and its context. Do not silently expand the requested scope.

Make it your own

User Request -> Orchestrator -> Language Model
Orchestrator -> Tool Router -> Search
Orchestrator -> Memory
Orchestrator -> Response

Download connections

Checking workspace availability. You can explore and copy this example now.

A map of responsibilities

This AI agent architecture diagram separates coordination, generation, tool access, and memory. Selecting a component explains its role. The arrows are an authored system map; they do not describe every turn in an agent loop.

Workflow or agent?

A fixed sequence of model and tool calls can be a workflow. A system in which the model selects its next actions needs additional controls around permissions, retries, and stopping. Use this example as a starting point, then label the decisions your implementation actually makes.

Three useful refinements

Add a human review component before actions that need approval. Separate read-only tools from tools that change external state. Give retries a bounded route rather than drawing an unlimited loop. Archifyi helps explain this design; the example does not execute an AI agent.

Further reading: Anthropic: Building effective agents ↗. The example above is independently authored and simplified.