Design a Single-Loop Agent

About this workflow

Build the core agent loop the leading coding agents use: one model, one loop, tools in, results folded back into context.

What happens when you run it

1. 1
   Write down exactly what one turn of the loop does before coding it.

   Runs flow-design-single-loop-agent-1-define-the-loop-contract on Claude Sonnet 4.6
2. 2
   Tools are the agent's hands - define them with strict schemas.

   Runs flow-design-single-loop-agent-2-build-the-tool-registry on Claude Sonnet 4.6
3. 3
   The loop's job is bookkeeping: order, roles, and IDs must be exact.

   Runs flow-design-single-loop-agent-3-implement-the-loop-with-result-folding on Claude Sonnet 4.6
4. 4
   An agent that cannot stop is worse than one that cannot start.

   Runs flow-design-single-loop-agent-4-harden-termination-and-error-paths on Claude Sonnet 4.6
5. 5
   Judge the loop the way harness engineers do: by reading transcripts.

   Runs flow-design-single-loop-agent-5-prove-it-with-a-transcript-review on Claude Sonnet 4.6

Best for

• •Recurring tasks you currently copy-paste between tools.
• •When a single prompt isn't enough — you need a chain where output(n) feeds prompt(n+1).
• •Fork and adapt if your inputs differ materially from the author's use case.

Models used