Inner/Outer Harness Pattern

A design pattern for autonomous agents that explicitly separates the **inner harness** (the loop concerned with model behavior: prompt formatting, tool schema, response parsing, model choice, retry logic for individual LLM calls) from the **outer harness** (the loop concerned with infrastructure: durable execution, checkpoint persistence, failure recovery, suspension/resumption, observability across runs). Each harness has independent technology choices, independent change cadence, and independent operational concerns; the contract between them is a small set of step-boundary primitives.

Most early agent frameworks mixed both concerns — LangChain pre-2024 was a single layer doing prompt assembly *and* retry policy *and* state management *and* infrastructure binding. The result was that swapping the model required changing the runtime, and swapping the runtime required rewriting the agent. The harness split makes each axis independently changeable: you can swap Claude for GPT-5 without touching durable-execution code; you can move from a single-machine runtime to Kitaru without rewriting the agent's tool-calling logic.