> For the complete documentation index, see [llms.txt](/llms.txt).
> A full single-fetch corpus is available at [llms-full.txt](/llms-full.txt).
---
title: "The AGNT5 Loop"
description: "Build → Ship → Run → Observe → Improve. The production loop for reliable AI workflows."
page_type: explanation
audience: both
last_verified: 2026-05-13
---

Building reliable AI workflows usually means stitching together an orchestrator, an observability stack, an eval pipeline, and custom glue code. AGNT5 brings those layers into one production loop, so the workflow you ship, the run you debug, and the eval you use to improve it all share the same execution context.

<div class="agnt5-loop-diagram" role="img" aria-label="The AGNT5 loop: Build → Ship → Run → Observe → Improve, with Improve feeding back into Build">
  <div class="agnt5-loop-track">
    <div class="agnt5-loop-stage">
      <span class="agnt5-loop-num">01</span>
      <span class="agnt5-loop-name">Build</span>
      <span class="agnt5-loop-note">Write a function in Python or TypeScript</span>
    </div>
    <span class="agnt5-loop-arrow" aria-hidden="true">→</span>
    <div class="agnt5-loop-stage">
      <span class="agnt5-loop-num">02</span>
      <span class="agnt5-loop-name">Ship</span>
      <span class="agnt5-loop-note">Deploy to managed Cloud</span>
    </div>
    <span class="agnt5-loop-arrow" aria-hidden="true">→</span>
    <div class="agnt5-loop-stage">
      <span class="agnt5-loop-num">03</span>
      <span class="agnt5-loop-name">Run</span>
      <span class="agnt5-loop-note">Invoke from your app, CLI, or schedule</span>
    </div>
    <span class="agnt5-loop-arrow" aria-hidden="true">→</span>
    <div class="agnt5-loop-stage">
      <span class="agnt5-loop-num">04</span>
      <span class="agnt5-loop-name">Observe</span>
      <span class="agnt5-loop-note">Trace every step, retry, and state mutation</span>
    </div>
    <span class="agnt5-loop-arrow" aria-hidden="true">→</span>
    <div class="agnt5-loop-stage">
      <span class="agnt5-loop-num">05</span>
      <span class="agnt5-loop-name">Improve</span>
      <span class="agnt5-loop-note">Capture a failure, write an eval, fix</span>
    </div>
  </div>
  <div class="agnt5-loop-return" aria-hidden="true">
    <span class="agnt5-loop-return-label">closes the loop</span>
  </div>
</div>

The differentiator is not any single phase. Durable execution exists elsewhere. So do observability tools and eval platforms. What AGNT5 ships is the connection between them: traces are tied to real runs, evals can point back to the executions that produced them, and fixes can be compared against the failures they are meant to resolve.

<div class="agnt5-loop-cta-row">
  <a href="/docs/get-started/build" class="agnt5-loop-cta">Start with Build →</a>
</div>