Patterns

Agent Architecture Patterns

Practical engineering patterns for memory, RAG, MCP tool access, and evaluation-before-autonomy workflows.

Pattern

Retrieval Grounding Loop

A bounded loop where the agent retrieves evidence, checks sufficiency, and retrieves again only when needed.

Read pattern

Pattern

Explicit Long-Term Memory

A pattern where the product defines exactly which facts can become durable memory.

Read pattern

Pattern

MCP Tool Gateway

A protocol gateway that exposes a narrow, governed tool surface to multiple agent clients.

Read pattern

Pattern

Eval Before Autonomy

A release pattern where every new autonomous action is gated by traces and a focused eval set.

Read pattern