agentic-ai-engineering/site/.vitepress/dist/assets/blog_posts_three-x-rule.md.Dm6HtTS4.js

import{c as s,Q as e,j as i,m as a}from"./chunks/framework.BPKcPtvA.js";const u=JSON.parse('{"title":"The 3x Rule of Agent Costs","description":"","frontmatter":{},"headers":[],"relativePath":"blog/posts/three-x-rule.md","filePath":"blog/posts/three-x-rule.md","lastUpdated":1780488246000}'),n={name:"blog/posts/three-x-rule.md"};function l(h,t,r,o,p,d){return e(),i("div",null,[...t[0]||(t[0]=[a(`<h1 id="the-3x-rule-of-agent-costs" tabindex="-1">The 3x Rule of Agent Costs <a class="header-anchor" href="#the-3x-rule-of-agent-costs" aria-label="Permalink to &quot;The 3x Rule of Agent Costs&quot;">​</a></h1><p><strong>June 10, 2026</strong></p><p>Here is a rule that will save you from budget surprises: <strong>production agent costs 3x your prototype estimate.</strong></p><h2 id="why-the-multiplier-exists" tabindex="-1">Why the Multiplier Exists <a class="header-anchor" href="#why-the-multiplier-exists" aria-label="Permalink to &quot;Why the Multiplier Exists&quot;">​</a></h2><table tabindex="0"><thead><tr><th>Phase</th><th>Multiplier</th><th>What Happens</th></tr></thead><tbody><tr><td>Prototype</td><td>1x</td><td>Happy path works perfectly</td></tr><tr><td>With retries</td><td>1.5x</td><td>Failed tool calls retry, edge cases handled</td></tr><tr><td>Production</td><td>3x</td><td>Monitoring, error handling, observability, security</td></tr></tbody></table><h2 id="where-the-cost-goes" tabindex="-1">Where the Cost Goes <a class="header-anchor" href="#where-the-cost-goes" aria-label="Permalink to &quot;Where the Cost Goes&quot;">​</a></h2><p>Output tokens dominate — about 70% of total cost. The model&#39;s reasoning is the expensive part. Input tokens (context) are about 20%. Cached tokens are 10%.</p><p>Every tool call costs 3-5x more than the call itself:</p><ul><li>Planning which tool to use</li><li>Executing the tool</li><li>Error recovery if it fails</li><li>Parsing the result</li><li>Adding the result back to context</li></ul><h2 id="quick-estimation" tabindex="-1">Quick Estimation <a class="header-anchor" href="#quick-estimation" aria-label="Permalink to &quot;Quick Estimation&quot;">​</a></h2><div class="language-python vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">python</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">def</span><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;"> estimate_cost</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">(turns, tokens_per_turn, price_per_m):</span></span>
<span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">    base </span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">=</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> turns </span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">*</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> tokens_per_turn </span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">*</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> price_per_m </span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">/</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> 1_000_000</span></span>
<span class="line"><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">    return</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> {</span></span>
<span class="line"><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">        &quot;prototype&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: base,</span></span>
<span class="line"><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">        &quot;with_retries&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: base </span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">*</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> 1.5</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">,</span></span>
<span class="line"><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">        &quot;production&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: base </span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">*</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> 3.0</span></span>
<span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">    }</span></span></code></pre></div><h2 id="budget-accordingly" tabindex="-1">Budget Accordingly <a class="header-anchor" href="#budget-accordingly" aria-label="Permalink to &quot;Budget Accordingly&quot;">​</a></h2><p>If your prototype agent costs $0.10 per task, plan for $0.30 in production. At 10,000 tasks per month, that is $3,000 per month, not $1,000. The 3x rule keeps you honest.</p><hr><p><em>From Module 6 of the <a href="/">Agentic Engineering Course</a>. The full module covers cost optimization, cascade routing, and pass@k evaluation.</em></p>`,15)])])}const c=s(n,[["render",l]]);export{u as __pageData,c as default};