AI Agent Pattern Catalog

Interleaves reasoning traces with tool actions in a Thought → Action → Observation loop until the task is complete.

A Planner LLM decomposes complex tasks into a structured plan; separate Executor agents carry out each step in parallel.

A lightweight classifier routes incoming requests to the most appropriate specialized sub-agent or workflow.

An orchestrator agent manages a fleet of specialized sub-agents, assigning tasks, monitoring progress, and aggregating outputs.

The agent pauses at defined checkpoints and requests human approval before executing high-risk, irreversible, or ambiguous actions.

Retrieves relevant documents from a vector store before each LLM call to ground responses in factual, up-to-date context.

Autonomously writes, executes, debugs, and iterates on code to complete software engineering tasks.

Maintains persistent memory across sessions using short-term context, episodic memory, and semantic vector memory.

Controls a web browser to navigate, extract information, fill forms, and interact with any website as a human would.

Uses the LLM's native function-calling capability to invoke structured tools and return typed results.