Skip to content

an open-source alternative to the dots, bots and muses of the world, run an office of claude code/codex like agents on your laptop, sandboxes or anywhere, uses your existing subscriptions

blog: the-multi-agent-cost-playbook (75 → 76)

Chaitanya Giria5ce63bcdd4d
diff --git a/blog/src/posts/the-multi-agent-cost-playbook.md b/blog/src/posts/the-multi-agent-cost-playbook.md
new file mode 100644
index 00000000..8f298322
--- /dev/null
+++ b/blog/src/posts/the-multi-agent-cost-playbook.md
@@ -0,0 +1,131 @@
+---
+title: "The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend"
+description: "Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet's bill, and how they compound."
+date: 2026-06-05
+category: guides
+categoryLabel: Guides
+type: Technical
+primaryKeyword: "ai agent cost optimization"
+secondaryKeywords: ["multi-agent cost", "llm cost optimization", "prompt caching", "model routing"]
+tags: ["Cost", "Model Routing", "Prompt Caching", "Multi-Agent"]
+author:
+ name: Chaitanya Giri
+ initials: CG
+faq:
+ - q: "What's the single biggest lever for cutting AI agent costs?"
+ a: "Prompt caching, for most agents. An agent re-sends the same large stable prefix — system prompt, tool definitions, project context — on every turn. With caching, that prefix is written once and then read at roughly a tenth of the normal input price, so only the per-turn delta pays full rate. For a long-running agent that's the difference between linear and near-flat context cost."
+ - q: "Do these levers stack?"
+ a: "Yes, and that's the point. Caching cuts the cost of the repeated prefix, batching halves eligible non-urgent work, model tiering moves the routine majority to a cheap model, context discipline shrinks what each agent carries, and local-first removes the platform fee on top of tokens. They're multiplicative, not either-or — a fleet applies several at once."
+ - q: "Does using cheaper models hurt quality?"
+ a: "Only if you route the wrong tasks to them. Most agent work is routine and a small model handles it fine; you reserve the frontier model for the genuinely hard minority. The skill is matching task difficulty to model tier — default lean, escalate on signal — not flattening everything onto one model."
+ - q: "How does a multi-agent harness help with cost specifically?"
+ a: "It gives you the structural levers in one place: per-agent model selection (cheap workers, premium orchestrator), a way to batch non-urgent fleet work, and — if it's local-first — no per-seat platform tax on top of model tokens. You pay for tokens, not for orchestration."
+---
+
+<div class="callout tldr"><span class="ic">TL;DR</span><p>Cost is the quiet tax on multi-agent
+systems: one agent's token rate × N agents, each carrying its own context, becomes a real bill fast. The
+good news is the biggest levers are mechanical and <strong>compound</strong>. Five of them:
+<strong>(1) prompt caching</strong> (reuse the stable prefix at ~a tenth the price),
+<strong>(2) batching</strong> (~50% off non-urgent work), <strong>(3) model tiering</strong> (cheap
+workers, premium orchestrator), <strong>(4) context discipline</strong> (don't pay to re-read
+everything, N times), and <strong>(5) local-first</strong> (no platform tax on top of tokens). Pull all
+five and a fleet's bill drops by more than half without losing quality.</p></div>
+
+Two earlier posts covered single levers: [routing the right task to the right
+model](/blog/do-more-with-less-model-routing/) and [not paying twice for the same
+tokens](/blog/prompt-caching-for-ai-agents/). This is the unified playbook — the **five levers** that
+actually move a multi-agent fleet's bill, why they matter more when agents run in parallel, and how they
+stack.
+
+Why multi-agent makes cost urgent: a single agent's token rate, multiplied by N agents running at once,
+each carrying its **own** context window, adds up fast. The flip side is that the levers are
+well-documented and composable — so the same parallelism that grows the bill is what makes disciplined
+optimization pay off N times over.
+
+## Lever 1 — Prompt caching: reuse the stable context
+
+The highest-leverage move for most agents, because an agent re-sends a large **stable** prefix every
+turn: its system prompt, tool definitions, and project/codebase context. Without caching you re-pay for
+all of it on every call.
+
+The mechanics are provider-documented: a cache **write** costs a small premium over base input (on
+Anthropic, ~1.25× for the short TTL), and every cache **read** costs roughly **a tenth** of the normal
+input price — about a 90% discount on the cached portion. Put the unchanging prefix first, cache it, and
+only the per-turn delta pays full rate. For a [long-running agent](/blog/long-running-agents-the-2026-shift/)
+that's the difference between linear and near-flat context cost. One caveat worth knowing: caches have a
+short time-to-live, so a bursty or idle agent can lose the cache between turns — pacing matters.
+
+## Lever 2 — Batching: trade latency for ~50%
+
+Provider **batch** APIs run requests asynchronously (results within ~24h) for about **50% off** input
+and output, with no quality difference — only timing. The agent insight is that a lot of fleet work
+isn't latency-sensitive at all: overnight reviews, bulk file analysis, [scheduled
+audits](/blog/scheduling-autonomous-agent-missions/). That work is batchable by design. Caching and
+batching together stack on eligible workloads for a large combined discount.
+
+## Lever 3 — Model tiering: cheap workers, premium orchestrator
+
+This is the structural multi-agent win. The price spread between a small model and a frontier model is
+enormous — on the order of **10–100× per token** — so running every agent on the biggest model is pure
+waste. The pattern: **route the routine majority to a cheap model and reserve the frontier model for the
+hard minority** (and for the orchestrator that has to reason about the whole job). Most agent work is
+routine, so this alone typically cuts spend by more than half without hurting quality. The
+[full argument is here](/blog/do-more-with-less-model-routing/).
+
+This is a concrete feature in Munder Difflin, not just advice: the harness's per-agent model selection
+(`HarnessConfig.defaultModel` in `src/main/config.ts`) lets you assign a Haiku-class model to worker
+agents and an Opus-class model to the GOD orchestrator **today** — the cheap-workers/premium-lead pattern
+configured directly.
+
+## Lever 4 — Context discipline: don't pay to re-read everything, N times
+
+The multiplier nobody budgets for: every agent's context window is a **recurring per-turn cost**, and in
+a fleet you pay it N times over. Keeping each agent's context **scoped** to only what its task needs,
+summarizing or compacting long histories, and not broadcasting full shared state into every agent's
+window are direct cost levers — not just hygiene. Cheaper recall is another reason a hive leans on
+[compact, markdown-first memory](/blog/markdown-first-agent-memory/) and semantic recall instead of
+dragging everything back into context.
+
+## Lever 5 — Local-first: remove the platform tax
+
+The structural one. Cloud agent *platforms* often charge per-seat or orchestration fees **on top of**
+the model token cost. A [local-first hive](/blog/why-local-first-matters-for-ai-agents/) pays **only**
+for model tokens — the orchestration, the files, the audit log all run on your machine for free. Pair
+that with an in-app usage view ([observability](/blog/observability-for-agent-fleets/)) and you can watch
+the bill climb in real time and apply levers 1–4 exactly where they bite. Predictable, attributable cost
+is part of the local-first case.
+
+## They compound — that's the whole point
+
+These aren't five alternatives; they're five multipliers you apply at once. Cache the stable prefix,
+batch the non-urgent work, tier the models, scope the context, and skip the platform tax. Each one bends
+the curve, and together they turn "we can't afford a fleet" into "the fleet costs less than the one
+over-powered agent it replaced." And remember the real denominator: not cost per token but **cost per
+completed task** — because a "cheap per token" agent that needs three retries to finish isn't cheap at
+all.
+
+## FAQ
+
+**What's the single biggest lever?** Prompt caching, for most agents — the stable prefix (system prompt,
+tools, context) gets written once and read at ~a tenth the price, so only the per-turn delta pays full
+rate.
+
+**Do these levers stack?** Yes — they're multiplicative. Caching cuts the repeated prefix, batching
+halves eligible non-urgent work, tiering moves the routine majority to a cheap model, context discipline
+shrinks each window, and local-first removes the platform fee. A fleet applies several at once.
+
+**Does cheaper hurt quality?** Only if you route the wrong tasks to a small model. Default lean, escalate
+the genuinely hard minority — match difficulty to tier rather than flattening everything onto one model.
+
+## The bottom line
+
+A multi-agent fleet doesn't have to be expensive — it has to be *engineered*. The five levers are
+mechanical, documented, and compounding: caching, batching, tiering, context discipline, and local-first.
+Pull them together and you spend less, run faster, and stop paying frontier prices for routine work.
+
+Munder Difflin is built for this: [per-agent model selection, local-first execution, and a usage view
+you can actually watch](https://munderdiffl.in/#how). [Download Munder Difflin](https://munderdiffl.in/#install)
+to run a fleet that doesn't bankrupt you — it's free and open source.
+
+Sources: [Anthropic — Prompt caching](https://platform.claude.com/docs/en/build-with-claude/prompt-caching);
+[Anthropic — Pricing (batch & caching)](https://platform.claude.com/docs/en/about-claude/pricing).
diff --git a/docs/blog/agent-security-and-sandboxing/index.html b/docs/blog/agent-security-and-sandboxing/index.html
index 8890c686..84002fec 100644
--- a/docs/blog/agent-security-and-sandboxing/index.html
+++ b/docs/blog/agent-security-and-sandboxing/index.html
@@ -266,7 +266,7 @@ git, lifecycle hooks, and an audit trail — so safe-by-default is the path of l
<footer class="post-foot wrap"><div class="prevnext">
<a class="pn prev" href="/blog/agents-rule-of-two/"><span class="k">← Older</span><span class="t">The Agents Rule of Two: A Simple Safety Rule for Coding Agents</span></a>
- <a class="pn next" href="/blog/mcp-security-tool-poisoning/"><span class="k">Newer →</span><span class="t">MCP Tool Poisoning: How to Secure the Model Context Protocol</span></a>
+ <a class="pn next" href="/blog/the-multi-agent-cost-playbook/"><span class="k">Newer →</span><span class="t">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</span></a>
</div>
<section class="related">
<h2>Related in Guides</h2>
@@ -308,20 +308,20 @@ git, lifecycle hooks, and an audit trail — so safe-by-default is the path of l
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/building-reliable-ai-agents/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
- <span>6 min</span>
+ <span>5 min</span>
</div>
- <h3><a href="/blog/building-reliable-ai-agents/">Building Reliable AI Agents: Failure Is the Default</a></h3>
- <p class="dek">Agents fail constantly: tools error, steps stall, output goes wrong. Reliability isn&#39;t avoiding failure — it&#39;s making each one contained and recoverable.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/building-reliable-ai-agents/" aria-label="Read: Building Reliable AI Agents: Failure Is the Default">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/building-reliable-ai-agents/index.html b/docs/blog/building-reliable-ai-agents/index.html
index eb6ccdff..802f4462 100644
--- a/docs/blog/building-reliable-ai-agents/index.html
+++ b/docs/blog/building-reliable-ai-agents/index.html
@@ -339,20 +339,20 @@ by design; it’s free and open source.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/claude-code-automation-while-you-sleep/index.html b/docs/blog/claude-code-automation-while-you-sleep/index.html
index dcf55689..c80a3f4d 100644
--- a/docs/blog/claude-code-automation-while-you-sleep/index.html
+++ b/docs/blog/claude-code-automation-while-you-sleep/index.html
@@ -327,20 +327,20 @@ while you sleep; it’s free and open source.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/claude-code-multi-agent-setup-tutorial/index.html b/docs/blog/claude-code-multi-agent-setup-tutorial/index.html
index 61556c2f..ae5f7c85 100644
--- a/docs/blog/claude-code-multi-agent-setup-tutorial/index.html
+++ b/docs/blog/claude-code-multi-agent-setup-tutorial/index.html
@@ -316,20 +316,20 @@ coordinated team of Claude Code agents — it’s free, local, and open source.<
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/claude-code-subagents-vs-multi-agent-harness/index.html b/docs/blog/claude-code-subagents-vs-multi-agent-harness/index.html
index 5925678c..48847bc1 100644
--- a/docs/blog/claude-code-subagents-vs-multi-agent-harness/index.html
+++ b/docs/blog/claude-code-subagents-vs-multi-agent-harness/index.html
@@ -356,20 +356,20 @@ open source.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/context-engineering-for-ai-agents/index.html b/docs/blog/context-engineering-for-ai-agents/index.html
index 6881e4d7..0335bf5e 100644
--- a/docs/blog/context-engineering-for-ai-agents/index.html
+++ b/docs/blog/context-engineering-for-ai-agents/index.html
@@ -344,20 +344,20 @@ it’s free and open source.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/debugging-multi-agent-systems/index.html b/docs/blog/debugging-multi-agent-systems/index.html
index 0872e123..379ca575 100644
--- a/docs/blog/debugging-multi-agent-systems/index.html
+++ b/docs/blog/debugging-multi-agent-systems/index.html
@@ -307,20 +307,20 @@ to run agents you can actually trace; it’s free and open source.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/evaluating-ai-agent-reliability/index.html b/docs/blog/evaluating-ai-agent-reliability/index.html
index 521b8fed..469b1e5d 100644
--- a/docs/blog/evaluating-ai-agent-reliability/index.html
+++ b/docs/blog/evaluating-ai-agent-reliability/index.html
@@ -310,20 +310,20 @@ your bar.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/feed.xml b/docs/blog/feed.xml
index e655a209..957aaa81 100644
--- a/docs/blog/feed.xml
+++ b/docs/blog/feed.xml
@@ -31,6 +31,14 @@
<updated>2026-06-05T00:00:00.000Z</updated><category term="Guides" />
<summary>How the Model Context Protocol gets attacked — tool poisoning, rug pulls, command injection — and the layered defenses that secure your MCP tools.</summary>
</entry>
+ <entry>
+ <title>The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</title>
+ <link href="https://munderdiffl.in/blog/the-multi-agent-cost-playbook/" />
+ <id>https://munderdiffl.in/blog/the-multi-agent-cost-playbook/</id>
+ <published>2026-06-05T00:00:00.000Z</published>
+ <updated>2026-06-05T00:00:00.000Z</updated><category term="Guides" />
+ <summary>Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</summary>
+ </entry>
<entry>
<title>Running AI Agents Safely: Permission Modes and Sandboxing</title>
<link href="https://munderdiffl.in/blog/agent-security-and-sandboxing/" />
@@ -479,12 +487,4 @@
<updated>2026-06-02T00:00:00.000Z</updated><category term="Internals" />
<summary>How we render AI agents as avatars on a Pixi.js office floor — driven by real hook and message events, with seat assignment, pathing, and flying envelopes.</summary>
</entry>
- <entry>
- <title>What Are Claude Code Agents? And How to Use Many at Once</title>
- <link href="https://munderdiffl.in/blog/what-are-claude-code-agents/" />
- <id>https://munderdiffl.in/blog/what-are-claude-code-agents/</id>
- <published>2026-06-02T00:00:00.000Z</published>
- <updated>2026-06-02T00:00:00.000Z</updated><category term="Concepts" />
- <summary>Claude Code agents explained in plain English — what an agent actually is, how subagents differ, and the leap from one agent to a coordinated team.</summary>
- </entry>
</feed>
diff --git a/docs/blog/how-to-install-and-use-munder-difflin/index.html b/docs/blog/how-to-install-and-use-munder-difflin/index.html
index 78bfb75c..e9e9a43a 100644
--- a/docs/blog/how-to-install-and-use-munder-difflin/index.html
+++ b/docs/blog/how-to-install-and-use-munder-difflin/index.html
@@ -422,38 +422,38 @@ task.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/building-reliable-ai-agents/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
<time datetime="2026-06-04">Jun 4, 2026</time>
<span class="dot" aria-hidden="true"></span>
- <span>6 min</span>
+ <span>5 min</span>
</div>
- <h3><a href="/blog/building-reliable-ai-agents/">Building Reliable AI Agents: Failure Is the Default</a></h3>
- <p class="dek">Agents fail constantly: tools error, steps stall, output goes wrong. Reliability isn&#39;t avoiding failure — it&#39;s making each one contained and recoverable.</p>
+ <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
+ <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/building-reliable-ai-agents/" aria-label="Read: Building Reliable AI Agents: Failure Is the Default">Read →</a>
+ <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/how-to-run-multiple-claude-code-agents/index.html b/docs/blog/how-to-run-multiple-claude-code-agents/index.html
index 71217739..775cf22f 100644
--- a/docs/blog/how-to-run-multiple-claude-code-agents/index.html
+++ b/docs/blog/how-to-run-multiple-claude-code-agents/index.html
@@ -251,20 +251,20 @@ free and open source.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/human-in-the-loop-approving-ai-agents/index.html b/docs/blog/human-in-the-loop-approving-ai-agents/index.html
index d47f3d55..813faae1 100644
--- a/docs/blog/human-in-the-loop-approving-ai-agents/index.html
+++ b/docs/blog/human-in-the-loop-approving-ai-agents/index.html
@@ -307,20 +307,20 @@ you talk to? You can <a href="/#install">download Munder Difflin</a> free; it’
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/index.html b/docs/blog/index.html
index ed0201ea..ed944f33 100644
--- a/docs/blog/index.html
+++ b/docs/blog/index.html
@@ -85,9 +85,9 @@
building a self-coordinating hive of agents with a GOD orchestrator you talk to.</p>
<div class="filters" role="navigation" aria-label="Topics">
- <a class="chip" aria-current="page">All<span class="ct">75</span></a>
+ <a class="chip" aria-current="page">All<span class="ct">76</span></a>
- <a class="chip" href="/blog/topics/guides/">guides<span class="ct">20</span></a>
+ <a class="chip" href="/blog/topics/guides/">guides<span class="ct">21</span></a>
<a class="chip" href="/blog/topics/internals/">internals<span class="ct">14</span></a>
@@ -167,6 +167,24 @@
</div>
</div>
</article>
+<article class="card">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
+ <span>GUIDES</span>
+ </a>
+ <div class="body">
+ <div class="meta">
+ <time datetime="2026-06-05">Jun 5, 2026</time>
+ <span class="dot" aria-hidden="true"></span>
+ <span>5 min</span>
+ </div>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
+ <div class="foot">
+ <span class="tag kind">Guides</span>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
+ </div>
+ </div>
+</article>
<article class="card">
<a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
diff --git a/docs/blog/manage-multiple-claude-code-sessions/index.html b/docs/blog/manage-multiple-claude-code-sessions/index.html
index cfcfec37..466e9215 100644
--- a/docs/blog/manage-multiple-claude-code-sessions/index.html
+++ b/docs/blog/manage-multiple-claude-code-sessions/index.html
@@ -304,20 +304,20 @@ isolated agents, shared memory, and a single committer, all on your own machine.
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/mcp-and-skills-in-a-hive/index.html b/docs/blog/mcp-and-skills-in-a-hive/index.html
index afc370a2..c24f3e79 100644
--- a/docs/blog/mcp-and-skills-in-a-hive/index.html
+++ b/docs/blog/mcp-and-skills-in-a-hive/index.html
@@ -320,20 +320,20 @@ existing setup into a coordinated team of agents; it’s free and open source.</
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/mcp-security-tool-poisoning/index.html b/docs/blog/mcp-security-tool-poisoning/index.html
index 9f372821..3daf3865 100644
--- a/docs/blog/mcp-security-tool-poisoning/index.html
+++ b/docs/blog/mcp-security-tool-poisoning/index.html
@@ -266,7 +266,7 @@ execution. Do that and MCP is what it should be: a powerful, <em>bounded</em> wa
</div>
<footer class="post-foot wrap"><div class="prevnext">
- <a class="pn prev" href="/blog/agent-security-and-sandboxing/"><span class="k">← Older</span><span class="t">Running AI Agents Safely: Permission Modes and Sandboxing</span></a>
+ <a class="pn prev" href="/blog/the-multi-agent-cost-playbook/"><span class="k">← Older</span><span class="t">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</span></a>
<a class="pn next" href="/blog/how-to-install-and-use-munder-difflin/"><span class="k">Newer →</span><span class="t">How to Install and Use Munder Difflin</span></a>
</div>
<section class="related">
@@ -291,38 +291,38 @@ execution. Do that and MCP is what it should be: a powerful, <em>bounded</em> wa
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/building-reliable-ai-agents/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
<time datetime="2026-06-04">Jun 4, 2026</time>
<span class="dot" aria-hidden="true"></span>
- <span>6 min</span>
+ <span>5 min</span>
</div>
- <h3><a href="/blog/building-reliable-ai-agents/">Building Reliable AI Agents: Failure Is the Default</a></h3>
- <p class="dek">Agents fail constantly: tools error, steps stall, output goes wrong. Reliability isn&#39;t avoiding failure — it&#39;s making each one contained and recoverable.</p>
+ <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
+ <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/building-reliable-ai-agents/" aria-label="Read: Building Reliable AI Agents: Failure Is the Default">Read →</a>
+ <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/prompt-caching-for-ai-agents/index.html b/docs/blog/prompt-caching-for-ai-agents/index.html
index 49ba2fe1..264bf486 100644
--- a/docs/blog/prompt-caching-for-ai-agents/index.html
+++ b/docs/blog/prompt-caching-for-ai-agents/index.html
@@ -339,20 +339,20 @@ that doesn’t waste your tokens; it’s free and open source.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/security-for-ai-coding-agents/index.html b/docs/blog/security-for-ai-coding-agents/index.html
index d0ae3e1c..7f77772a 100644
--- a/docs/blog/security-for-ai-coding-agents/index.html
+++ b/docs/blog/security-for-ai-coding-agents/index.html
@@ -304,20 +304,20 @@ and open source.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/small-team-ai-agents-playbook/index.html b/docs/blog/small-team-ai-agents-playbook/index.html
index 1a58086b..60799de5 100644
--- a/docs/blog/small-team-ai-agents-playbook/index.html
+++ b/docs/blog/small-team-ai-agents-playbook/index.html
@@ -309,20 +309,20 @@ it’s free and open source.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/tags/cost/index.html b/docs/blog/tags/cost/index.html
index 53af02a4..15c3d30c 100644
--- a/docs/blog/tags/cost/index.html
+++ b/docs/blog/tags/cost/index.html
@@ -58,11 +58,29 @@
<header class="page-head wrap">
<p class="eyebrow">Tag</p>
<h1>#Cost</h1>
- <p class="lead">2 posts tagged <strong>Cost</strong>.</p>
+ <p class="lead">3 posts tagged <strong>Cost</strong>.</p>
</header>
<section class="wrap">
<div class="post-grid">
<article class="card">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
+ <span>GUIDES</span>
+ </a>
+ <div class="body">
+ <div class="meta">
+ <time datetime="2026-06-05">Jun 5, 2026</time>
+ <span class="dot" aria-hidden="true"></span>
+ <span>5 min</span>
+ </div>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
+ <div class="foot">
+ <span class="tag kind">Guides</span>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
+ </div>
+ </div>
+</article>
+<article class="card">
<a class="thumb t-concepts" href="/blog/do-more-with-less-model-routing/" aria-hidden="true" tabindex="-1">
<span>CONCEPTS</span>
</a>
diff --git a/docs/blog/tags/model-routing/index.html b/docs/blog/tags/model-routing/index.html
index 2990642a..01b8f21f 100644
--- a/docs/blog/tags/model-routing/index.html
+++ b/docs/blog/tags/model-routing/index.html
@@ -58,11 +58,29 @@
<header class="page-head wrap">
<p class="eyebrow">Tag</p>
<h1>#Model Routing</h1>
- <p class="lead">1 post tagged <strong>Model Routing</strong>.</p>
+ <p class="lead">2 posts tagged <strong>Model Routing</strong>.</p>
</header>
<section class="wrap">
<div class="post-grid">
<article class="card">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
+ <span>GUIDES</span>
+ </a>
+ <div class="body">
+ <div class="meta">
+ <time datetime="2026-06-05">Jun 5, 2026</time>
+ <span class="dot" aria-hidden="true"></span>
+ <span>5 min</span>
+ </div>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
+ <div class="foot">
+ <span class="tag kind">Guides</span>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
+ </div>
+ </div>
+</article>
+<article class="card">
<a class="thumb t-concepts" href="/blog/do-more-with-less-model-routing/" aria-hidden="true" tabindex="-1">
<span>CONCEPTS</span>
</a>
diff --git a/docs/blog/tags/multi-agent/index.html b/docs/blog/tags/multi-agent/index.html
index c3969f74..4d902d75 100644
--- a/docs/blog/tags/multi-agent/index.html
+++ b/docs/blog/tags/multi-agent/index.html
@@ -58,11 +58,29 @@
<header class="page-head wrap">
<p class="eyebrow">Tag</p>
<h1>#Multi-Agent</h1>
- <p class="lead">50 posts tagged <strong>Multi-Agent</strong>.</p>
+ <p class="lead">51 posts tagged <strong>Multi-Agent</strong>.</p>
</header>
<section class="wrap">
<div class="post-grid">
<article class="card">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
+ <span>GUIDES</span>
+ </a>
+ <div class="body">
+ <div class="meta">
+ <time datetime="2026-06-05">Jun 5, 2026</time>
+ <span class="dot" aria-hidden="true"></span>
+ <span>5 min</span>
+ </div>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
+ <div class="foot">
+ <span class="tag kind">Guides</span>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
+ </div>
+ </div>
+</article>
+<article class="card">
<a class="thumb t-concepts" href="/blog/agents-rule-of-two/" aria-hidden="true" tabindex="-1">
<span>CONCEPTS</span>
</a>
diff --git a/docs/blog/tags/prompt-caching/index.html b/docs/blog/tags/prompt-caching/index.html
index 65a08917..8a868b12 100644
--- a/docs/blog/tags/prompt-caching/index.html
+++ b/docs/blog/tags/prompt-caching/index.html
@@ -58,11 +58,29 @@
<header class="page-head wrap">
<p class="eyebrow">Tag</p>
<h1>#Prompt Caching</h1>
- <p class="lead">1 post tagged <strong>Prompt Caching</strong>.</p>
+ <p class="lead">2 posts tagged <strong>Prompt Caching</strong>.</p>
</header>
<section class="wrap">
<div class="post-grid">
<article class="card">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
+ <span>GUIDES</span>
+ </a>
+ <div class="body">
+ <div class="meta">
+ <time datetime="2026-06-05">Jun 5, 2026</time>
+ <span class="dot" aria-hidden="true"></span>
+ <span>5 min</span>
+ </div>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
+ <div class="foot">
+ <span class="tag kind">Guides</span>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
+ </div>
+ </div>
+</article>
+<article class="card">
<a class="thumb t-guides" href="/blog/prompt-caching-for-ai-agents/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
diff --git a/docs/blog/the-lethal-trifecta-for-coding-agents/index.html b/docs/blog/the-lethal-trifecta-for-coding-agents/index.html
index 5a6b4cd4..93a42d9c 100644
--- a/docs/blog/the-lethal-trifecta-for-coding-agents/index.html
+++ b/docs/blog/the-lethal-trifecta-for-coding-agents/index.html
@@ -327,20 +327,20 @@ secrets; it’s free and open source.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/the-multi-agent-cost-playbook/index.html b/docs/blog/the-multi-agent-cost-playbook/index.html
new file mode 100644
index 00000000..26013f6a
--- /dev/null
+++ b/docs/blog/the-multi-agent-cost-playbook/index.html
@@ -0,0 +1,371 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="UTF-8" />
+<meta name="viewport" content="width=device-width, initial-scale=1.0" />
+<title>The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend — Munder Difflin Blog</title>
+<meta name="description" content="Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound." />
+<link rel="canonical" href="https://munderdiffl.in/blog/the-multi-agent-cost-playbook/" />
+
+<link rel="icon" type="image/png" href="https://munderdiffl.in/logo.png" />
+<link rel="apple-touch-icon" href="https://munderdiffl.in/logo.png" />
+<meta name="theme-color" content="#F5F2E8" />
+
+<!-- Open Graph -->
+<meta property="og:site_name" content="Munder Difflin Blog" />
+<meta property="og:type" content="article" />
+<meta property="og:title" content="The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend — Munder Difflin Blog" />
+<meta property="og:description" content="Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound." />
+<meta property="og:url" content="https://munderdiffl.in/blog/the-multi-agent-cost-playbook/" />
+<meta property="og:image" content="https://munderdiffl.in/media/og.png" />
+<meta property="og:locale" content="en_US" />
+<meta property="article:published_time" content="2026-06-05T00:00:00.000Z" />
+<meta property="article:author" content="Chaitanya Giri" />
+<meta property="article:tag" content="Cost" />
+<meta property="article:tag" content="Model Routing" />
+<meta property="article:tag" content="Prompt Caching" />
+<meta property="article:tag" content="Multi-Agent" />
+
+
+<!-- Twitter -->
+<meta name="twitter:card" content="summary_large_image" />
+<meta name="twitter:title" content="The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend — Munder Difflin Blog" />
+<meta name="twitter:description" content="Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound." />
+<meta name="twitter:image" content="https://munderdiffl.in/media/og.png" />
+
+<link rel="alternate" type="application/atom+xml" title="Munder Difflin Blog" href="https://munderdiffl.in/blog/feed.xml" />
+<link rel="preconnect" href="https://fonts.googleapis.com" />
+<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin />
+<link href="https://fonts.googleapis.com/css2?family=Geist:wght@400;500;600&family=JetBrains+Mono:wght@400;500;600;700&display=swap" rel="stylesheet" />
+<link rel="stylesheet" href="/blog/assets/blog.css" />
+
+<script type="application/ld+json">
+{
+ "@context": "https://schema.org",
+ "@type": "BlogPosting",
+ "headline": "The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend",
+ "description": "Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet's bill, and how they compound.",
+ "image": ["https://munderdiffl.in/media/og.png"],
+ "datePublished": "2026-06-05T00:00:00.000Z",
+ "dateModified": "2026-06-05T00:00:00.000Z",
+ "author": { "@type": "Person", "name": "Chaitanya Giri" },
+ "publisher": {
+ "@type": "Organization",
+ "name": "Munder Difflin",
+ "logo": { "@type": "ImageObject", "url": "https://munderdiffl.in/logo.png" }
+ },
+ "mainEntityOfPage": { "@type": "WebPage", "@id": "https://munderdiffl.in/blog/the-multi-agent-cost-playbook/" },
+ "keywords": "ai agent cost optimization, multi-agent cost, llm cost optimization, prompt caching, model routing",
+ "articleSection": "Guides",
+ "inLanguage": "en-US",
+ "url": "https://munderdiffl.in/blog/the-multi-agent-cost-playbook/"
+}
+</script>
+<script type="application/ld+json">
+{
+ "@context": "https://schema.org",
+ "@type": "BreadcrumbList",
+ "itemListElement": [
+ { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://munderdiffl.in/" },
+ { "@type": "ListItem", "position": 2, "name": "Blog", "item": "https://munderdiffl.in/blog/" },
+ { "@type": "ListItem", "position": 3, "name": "The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend", "item": "https://munderdiffl.in/blog/the-multi-agent-cost-playbook/" }
+ ]
+}
+</script>
+
+<script type="application/ld+json">
+{
+ "@context": "https://schema.org",
+ "@type": "FAQPage",
+ "mainEntity": [
+ { "@type": "Question", "name": "What's the single biggest lever for cutting AI agent costs?", "acceptedAnswer": { "@type": "Answer", "text": "Prompt caching, for most agents. An agent re-sends the same large stable prefix — system prompt, tool definitions, project context — on every turn. With caching, that prefix is written once and then read at roughly a tenth of the normal input price, so only the per-turn delta pays full rate. For a long-running agent that's the difference between linear and near-flat context cost." } },
+ { "@type": "Question", "name": "Do these levers stack?", "acceptedAnswer": { "@type": "Answer", "text": "Yes, and that's the point. Caching cuts the cost of the repeated prefix, batching halves eligible non-urgent work, model tiering moves the routine majority to a cheap model, context discipline shrinks what each agent carries, and local-first removes the platform fee on top of tokens. They're multiplicative, not either-or — a fleet applies several at once." } },
+ { "@type": "Question", "name": "Does using cheaper models hurt quality?", "acceptedAnswer": { "@type": "Answer", "text": "Only if you route the wrong tasks to them. Most agent work is routine and a small model handles it fine; you reserve the frontier model for the genuinely hard minority. The skill is matching task difficulty to model tier — default lean, escalate on signal — not flattening everything onto one model." } },
+ { "@type": "Question", "name": "How does a multi-agent harness help with cost specifically?", "acceptedAnswer": { "@type": "Answer", "text": "It gives you the structural levers in one place: per-agent model selection (cheap workers, premium orchestrator), a way to batch non-urgent fleet work, and — if it's local-first — no per-seat platform tax on top of model tokens. You pay for tokens, not for orchestration." } }
+
+ ]
+}
+</script>
+
+
+</head>
+<body>
+<a class="skip-link" href="#main">Skip to content</a>
+<nav id="nav">
+ <a class="brand" href="/blog/" aria-label="Munder Difflin Blog home">
+ <span class="brand-mark">MD</span>
+ <span class="brand-word">MUNDER DIFFLIN<span class="sep">/</span><span class="blog">BLOG</span></span>
+ </a>
+ <span class="spacer"></span>
+ <div class="links">
+ <a href="/blog/">Latest</a>
+ <a href="/blog/topics/">Topics</a>
+ <a href="/blog/about/">About</a>
+ <a href="https://munderdiffl.in">↗ Product</a>
+ </div>
+ <div class="nav-actions">
+ <a class="btn sm" href="https://munderdiffl.in/blog/feed.xml" aria-label="RSS feed">⤵ RSS</a>
+ <a class="btn sm primary" href="https://munderdiffl.in/#install">⤓ Download</a>
+ </div>
+</nav>
+
+<main id="main">
+
+<article>
+ <header class="post-head wrap">
+ <nav class="crumbs" aria-label="Breadcrumb">
+ <a href="/blog/">Blog</a>
+ <span>/</span>
+ <a href="/blog/topics/guides/">Guides</a>
+ </nav>
+ <h1>The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</h1>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
+ <div class="byline">
+ <span class="av" aria-hidden="true">CG</span>
+ <strong>Chaitanya Giri</strong>
+ <span class="dot" aria-hidden="true">·</span>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
+ <span class="dot" aria-hidden="true">·</span>
+ <span>5 min read</span>
+ <span class="tag kind">Guides</span>
+ </div>
+ </header>
+
+ <div class="post-wrap wrap">
+ <div class="prose">
+ <div class="callout tldr"><span class="ic">TL;DR</span><p>Cost is the quiet tax on multi-agent
+systems: one agent's token rate × N agents, each carrying its own context, becomes a real bill fast. The
+good news is the biggest levers are mechanical and <strong>compound</strong>. Five of them:
+<strong>(1) prompt caching</strong> (reuse the stable prefix at ~a tenth the price),
+<strong>(2) batching</strong> (~50% off non-urgent work), <strong>(3) model tiering</strong> (cheap
+workers, premium orchestrator), <strong>(4) context discipline</strong> (don't pay to re-read
+everything, N times), and <strong>(5) local-first</strong> (no platform tax on top of tokens). Pull all
+five and a fleet's bill drops by more than half without losing quality.</p></div>
+<p>Two earlier posts covered single levers: <a href="/blog/do-more-with-less-model-routing/">routing the right task to the right
+model</a> and <a href="/blog/prompt-caching-for-ai-agents/">not paying twice for the same
+tokens</a>. This is the unified playbook — the <strong>five levers</strong> that
+actually move a multi-agent fleet’s bill, why they matter more when agents run in parallel, and how they
+stack.</p>
+<p>Why multi-agent makes cost urgent: a single agent’s token rate, multiplied by N agents running at once,
+each carrying its <strong>own</strong> context window, adds up fast. The flip side is that the levers are
+well-documented and composable — so the same parallelism that grows the bill is what makes disciplined
+optimization pay off N times over.</p>
+<h2 id="lever-1-prompt-caching-reuse-the-stable-context" tabindex="-1">Lever 1 — Prompt caching: reuse the stable context <a class="anchor" href="#lever-1-prompt-caching-reuse-the-stable-context" aria-hidden="true">#</a></h2>
+<p>The highest-leverage move for most agents, because an agent re-sends a large <strong>stable</strong> prefix every
+turn: its system prompt, tool definitions, and project/codebase context. Without caching you re-pay for
+all of it on every call.</p>
+<p>The mechanics are provider-documented: a cache <strong>write</strong> costs a small premium over base input (on
+Anthropic, ~1.25× for the short TTL), and every cache <strong>read</strong> costs roughly <strong>a tenth</strong> of the normal
+input price — about a 90% discount on the cached portion. Put the unchanging prefix first, cache it, and
+only the per-turn delta pays full rate. For a <a href="/blog/long-running-agents-the-2026-shift/">long-running agent</a>
+that’s the difference between linear and near-flat context cost. One caveat worth knowing: caches have a
+short time-to-live, so a bursty or idle agent can lose the cache between turns — pacing matters.</p>
+<h2 id="lever-2-batching-trade-latency-for-50" tabindex="-1">Lever 2 — Batching: trade latency for ~50% <a class="anchor" href="#lever-2-batching-trade-latency-for-50" aria-hidden="true">#</a></h2>
+<p>Provider <strong>batch</strong> APIs run requests asynchronously (results within ~24h) for about <strong>50% off</strong> input
+and output, with no quality difference — only timing. The agent insight is that a lot of fleet work
+isn’t latency-sensitive at all: overnight reviews, bulk file analysis, <a href="/blog/scheduling-autonomous-agent-missions/">scheduled
+audits</a>. That work is batchable by design. Caching and
+batching together stack on eligible workloads for a large combined discount.</p>
+<h2 id="lever-3-model-tiering-cheap-workers-premium-orchestrator" tabindex="-1">Lever 3 — Model tiering: cheap workers, premium orchestrator <a class="anchor" href="#lever-3-model-tiering-cheap-workers-premium-orchestrator" aria-hidden="true">#</a></h2>
+<p>This is the structural multi-agent win. The price spread between a small model and a frontier model is
+enormous — on the order of <strong>10–100× per token</strong> — so running every agent on the biggest model is pure
+waste. The pattern: <strong>route the routine majority to a cheap model and reserve the frontier model for the
+hard minority</strong> (and for the orchestrator that has to reason about the whole job). Most agent work is
+routine, so this alone typically cuts spend by more than half without hurting quality. The
+<a href="/blog/do-more-with-less-model-routing/">full argument is here</a>.</p>
+<p>This is a concrete feature in Munder Difflin, not just advice: the harness’s per-agent model selection
+(<code>HarnessConfig.defaultModel</code> in <code>src/main/config.ts</code>) lets you assign a Haiku-class model to worker
+agents and an Opus-class model to the GOD orchestrator <strong>today</strong> — the cheap-workers/premium-lead pattern
+configured directly.</p>
+<h2 id="lever-4-context-discipline-dont-pay-to-re-read-everything-n-times" tabindex="-1">Lever 4 — Context discipline: don’t pay to re-read everything, N times <a class="anchor" href="#lever-4-context-discipline-dont-pay-to-re-read-everything-n-times" aria-hidden="true">#</a></h2>
+<p>The multiplier nobody budgets for: every agent’s context window is a <strong>recurring per-turn cost</strong>, and in
+a fleet you pay it N times over. Keeping each agent’s context <strong>scoped</strong> to only what its task needs,
+summarizing or compacting long histories, and not broadcasting full shared state into every agent’s
+window are direct cost levers — not just hygiene. Cheaper recall is another reason a hive leans on
+<a href="/blog/markdown-first-agent-memory/">compact, markdown-first memory</a> and semantic recall instead of
+dragging everything back into context.</p>
+<h2 id="lever-5-local-first-remove-the-platform-tax" tabindex="-1">Lever 5 — Local-first: remove the platform tax <a class="anchor" href="#lever-5-local-first-remove-the-platform-tax" aria-hidden="true">#</a></h2>
+<p>The structural one. Cloud agent <em>platforms</em> often charge per-seat or orchestration fees <strong>on top of</strong>
+the model token cost. A <a href="/blog/why-local-first-matters-for-ai-agents/">local-first hive</a> pays <strong>only</strong>
+for model tokens — the orchestration, the files, the audit log all run on your machine for free. Pair
+that with an in-app usage view (<a href="/blog/observability-for-agent-fleets/">observability</a>) and you can watch
+the bill climb in real time and apply levers 1–4 exactly where they bite. Predictable, attributable cost
+is part of the local-first case.</p>
+<h2 id="they-compound-thats-the-whole-point" tabindex="-1">They compound — that’s the whole point <a class="anchor" href="#they-compound-thats-the-whole-point" aria-hidden="true">#</a></h2>
+<p>These aren’t five alternatives; they’re five multipliers you apply at once. Cache the stable prefix,
+batch the non-urgent work, tier the models, scope the context, and skip the platform tax. Each one bends
+the curve, and together they turn “we can’t afford a fleet” into “the fleet costs less than the one
+over-powered agent it replaced.” And remember the real denominator: not cost per token but <strong>cost per
+completed task</strong> — because a “cheap per token” agent that needs three retries to finish isn’t cheap at
+all.</p>
+<h2 id="faq" tabindex="-1">FAQ <a class="anchor" href="#faq" aria-hidden="true">#</a></h2>
+<p><strong>What’s the single biggest lever?</strong> Prompt caching, for most agents — the stable prefix (system prompt,
+tools, context) gets written once and read at ~a tenth the price, so only the per-turn delta pays full
+rate.</p>
+<p><strong>Do these levers stack?</strong> Yes — they’re multiplicative. Caching cuts the repeated prefix, batching
+halves eligible non-urgent work, tiering moves the routine majority to a cheap model, context discipline
+shrinks each window, and local-first removes the platform fee. A fleet applies several at once.</p>
+<p><strong>Does cheaper hurt quality?</strong> Only if you route the wrong tasks to a small model. Default lean, escalate
+the genuinely hard minority — match difficulty to tier rather than flattening everything onto one model.</p>
+<h2 id="the-bottom-line" tabindex="-1">The bottom line <a class="anchor" href="#the-bottom-line" aria-hidden="true">#</a></h2>
+<p>A multi-agent fleet doesn’t have to be expensive — it has to be <em>engineered</em>. The five levers are
+mechanical, documented, and compounding: caching, batching, tiering, context discipline, and local-first.
+Pull them together and you spend less, run faster, and stop paying frontier prices for routine work.</p>
+<p>Munder Difflin is built for this: <a href="https://munderdiffl.in/#how">per-agent model selection, local-first execution, and a usage view
+you can actually watch</a>. <a href="https://munderdiffl.in/#install">Download Munder Difflin</a>
+to run a fleet that doesn’t bankrupt you — it’s free and open source.</p>
+<p>Sources: <a href="https://platform.claude.com/docs/en/build-with-claude/prompt-caching">Anthropic — Prompt caching</a>;
+<a href="https://platform.claude.com/docs/en/about-claude/pricing">Anthropic — Pricing (batch &amp; caching)</a>.</p>
+
+
+
+ <section class="faq" aria-label="Frequently asked questions">
+ <h2 id="faq">FAQ</h2>
+
+ <div class="qa">
+ <p class="q">What&#39;s the single biggest lever for cutting AI agent costs?</p>
+ <p class="a">Prompt caching, for most agents. An agent re-sends the same large stable prefix — system prompt, tool definitions, project context — on every turn. With caching, that prefix is written once and then read at roughly a tenth of the normal input price, so only the per-turn delta pays full rate. For a long-running agent that&#39;s the difference between linear and near-flat context cost.</p>
+ </div>
+
+ <div class="qa">
+ <p class="q">Do these levers stack?</p>
+ <p class="a">Yes, and that&#39;s the point. Caching cuts the cost of the repeated prefix, batching halves eligible non-urgent work, model tiering moves the routine majority to a cheap model, context discipline shrinks what each agent carries, and local-first removes the platform fee on top of tokens. They&#39;re multiplicative, not either-or — a fleet applies several at once.</p>
+ </div>
+
+ <div class="qa">
+ <p class="q">Does using cheaper models hurt quality?</p>
+ <p class="a">Only if you route the wrong tasks to them. Most agent work is routine and a small model handles it fine; you reserve the frontier model for the genuinely hard minority. The skill is matching task difficulty to model tier — default lean, escalate on signal — not flattening everything onto one model.</p>
+ </div>
+
+ <div class="qa">
+ <p class="q">How does a multi-agent harness help with cost specifically?</p>
+ <p class="a">It gives you the structural levers in one place: per-agent model selection (cheap workers, premium orchestrator), a way to batch non-urgent fleet work, and — if it&#39;s local-first — no per-seat platform tax on top of model tokens. You pay for tokens, not for orchestration.</p>
+ </div>
+
+ </section>
+
+
+
+ <hr />
+ <div class="share">
+ <span class="lbl">Tags</span>
+ <a class="tag" href="/blog/tags/cost/">Cost</a><a class="tag" href="/blog/tags/model-routing/">Model Routing</a><a class="tag" href="/blog/tags/prompt-caching/">Prompt Caching</a><a class="tag" href="/blog/tags/multi-agent/">Multi-Agent</a>
+ </div>
+
+ </div>
+
+
+ <aside class="toc-rail" aria-label="Table of contents">
+ <div class="tt">On this page</div>
+ <ol>
+ <li class="lvl-2"><a href="#lever-1-prompt-caching-reuse-the-stable-context">Lever 1 — Prompt caching: reuse the stable context</a></li><li class="lvl-2"><a href="#lever-2-batching-trade-latency-for-50">Lever 2 — Batching: trade latency for ~50%</a></li><li class="lvl-2"><a href="#lever-3-model-tiering-cheap-workers-premium-orchestrator">Lever 3 — Model tiering: cheap workers, premium orchestrator</a></li><li class="lvl-2"><a href="#lever-4-context-discipline-dont-pay-to-re-read-everything-n-times">Lever 4 — Context discipline: don’t pay to re-read everything, N times</a></li><li class="lvl-2"><a href="#lever-5-local-first-remove-the-platform-tax">Lever 5 — Local-first: remove the platform tax</a></li><li class="lvl-2"><a href="#they-compound-thats-the-whole-point">They compound — that’s the whole point</a></li><li class="lvl-2"><a href="#faq">FAQ</a></li><li class="lvl-2"><a href="#the-bottom-line">The bottom line</a></li>
+ </ol>
+ </aside>
+
+ </div>
+
+ <footer class="post-foot wrap"><div class="prevnext">
+ <a class="pn prev" href="/blog/agent-security-and-sandboxing/"><span class="k">← Older</span><span class="t">Running AI Agents Safely: Permission Modes and Sandboxing</span></a>
+ <a class="pn next" href="/blog/mcp-security-tool-poisoning/"><span class="k">Newer →</span><span class="t">MCP Tool Poisoning: How to Secure the Model Context Protocol</span></a>
+ </div>
+ <section class="related">
+ <h2>Related in Guides</h2>
+ <div class="post-grid">
+ <article class="card">
+ <a class="thumb t-guides" href="/blog/how-to-install-and-use-munder-difflin/" aria-hidden="true" tabindex="-1">
+ <span>GUIDES</span>
+ </a>
+ <div class="body">
+ <div class="meta">
+ <time datetime="2026-06-05">Jun 5, 2026</time>
+ <span class="dot" aria-hidden="true"></span>
+ <span>9 min</span>
+ </div>
+ <h3><a href="/blog/how-to-install-and-use-munder-difflin/">How to Install and Use Munder Difflin</a></h3>
+ <p class="dek">Install Munder Difflin on macOS, Windows, or Linux and put a hive of Claude Code agents to work on ambitious, long-horizon tasks — start to finish.</p>
+ <div class="foot">
+ <span class="tag kind">Guides</span>
+ <a class="more" href="/blog/how-to-install-and-use-munder-difflin/" aria-label="Read: How to Install and Use Munder Difflin">Read →</a>
+ </div>
+ </div>
+</article>
+<article class="card">
+ <a class="thumb t-guides" href="/blog/mcp-security-tool-poisoning/" aria-hidden="true" tabindex="-1">
+ <span>GUIDES</span>
+ </a>
+ <div class="body">
+ <div class="meta">
+ <time datetime="2026-06-05">Jun 5, 2026</time>
+ <span class="dot" aria-hidden="true"></span>
+ <span>5 min</span>
+ </div>
+ <h3><a href="/blog/mcp-security-tool-poisoning/">MCP Tool Poisoning: How to Secure the Model Context Protocol</a></h3>
+ <p class="dek">How the Model Context Protocol gets attacked — tool poisoning, rug pulls, command injection — and the layered defenses that secure your MCP tools.</p>
+ <div class="foot">
+ <span class="tag kind">Guides</span>
+ <a class="more" href="/blog/mcp-security-tool-poisoning/" aria-label="Read: MCP Tool Poisoning: How to Secure the Model Context Protocol">Read →</a>
+ </div>
+ </div>
+</article>
+<article class="card">
+ <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <span>GUIDES</span>
+ </a>
+ <div class="body">
+ <div class="meta">
+ <time datetime="2026-06-04">Jun 4, 2026</time>
+ <span class="dot" aria-hidden="true"></span>
+ <span>5 min</span>
+ </div>
+ <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
+ <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <div class="foot">
+ <span class="tag kind">Guides</span>
+ <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ </div>
+ </div>
+</article>
+
+ </div>
+ </section>
+
+ </footer>
+</article>
+
+</main>
+<footer>
+ <div class="wrap">
+ <div class="foot-top">
+ <a class="foot-brand brand" href="/blog/">
+ <span class="brand-mark">MD</span>
+ <span class="brand-word">MUNDER DIFFLIN<span class="sep">/</span><span class="blog">BLOG</span></span>
+ </a>
+ <div class="foot-links">
+ <a href="/blog/">Latest</a>
+ <a href="/blog/topics/">Topics</a>
+ <a href="/blog/about/">About</a>
+ <a href="https://munderdiffl.in/blog/feed.xml">RSS</a>
+ <a href="https://github.com/chaitanyagiri/munder-difflin">GitHub</a>
+ <a href="https://munderdiffl.in">munderdiffl.in ↗</a>
+ </div>
+ </div>
+ <div class="foot-meta">
+ © 2026 Munder Difflin · Local-first · MIT-licensed ·
+ Built on <a href="https://munderdiffl.in">Electron, React, Pixi.js, xterm.js &amp; node-pty</a>.
+ Run an office of agents while you sleep.
+ </div>
+ </div>
+</footer>
+
+<script>
+ // sticky-nav border on scroll
+ (function () {
+ var nav = document.getElementById('nav');
+ if (!nav) return;
+ var onScroll = function () { nav.classList.toggle('scrolled', window.scrollY > 8); };
+ onScroll(); window.addEventListener('scroll', onScroll, { passive: true });
+ })();
+</script>
+
+</body>
+</html>
diff --git a/docs/blog/topics/comparisons/index.html b/docs/blog/topics/comparisons/index.html
index 123f7f7e..1d342f75 100644
--- a/docs/blog/topics/comparisons/index.html
+++ b/docs/blog/topics/comparisons/index.html
@@ -68,7 +68,7 @@
<div class="filters" role="navigation" aria-label="Topics">
<a class="chip" href="/blog/">All</a>
- <a class="chip" href="/blog/topics/guides/">guides<span class="ct">20</span></a>
+ <a class="chip" href="/blog/topics/guides/">guides<span class="ct">21</span></a>
<a class="chip" href="/blog/topics/internals/">internals<span class="ct">14</span></a>
diff --git a/docs/blog/topics/concepts/index.html b/docs/blog/topics/concepts/index.html
index e0ae30fb..57268cee 100644
--- a/docs/blog/topics/concepts/index.html
+++ b/docs/blog/topics/concepts/index.html
@@ -68,7 +68,7 @@
<div class="filters" role="navigation" aria-label="Topics">
<a class="chip" href="/blog/">All</a>
- <a class="chip" href="/blog/topics/guides/">guides<span class="ct">20</span></a>
+ <a class="chip" href="/blog/topics/guides/">guides<span class="ct">21</span></a>
<a class="chip" href="/blog/topics/internals/">internals<span class="ct">14</span></a>
diff --git a/docs/blog/topics/guides/index.html b/docs/blog/topics/guides/index.html
index d7427081..b0a60d8a 100644
--- a/docs/blog/topics/guides/index.html
+++ b/docs/blog/topics/guides/index.html
@@ -63,12 +63,12 @@
</nav>
<p class="eyebrow">Topic</p>
<h1>guides</h1>
- <p class="lead">20 posts in this cluster.</p>
+ <p class="lead">21 posts in this cluster.</p>
<div class="filters" role="navigation" aria-label="Topics">
<a class="chip" href="/blog/">All</a>
- <a class="chip" href="/blog/topics/guides/" aria-current="page">guides<span class="ct">20</span></a>
+ <a class="chip" href="/blog/topics/guides/" aria-current="page">guides<span class="ct">21</span></a>
<a class="chip" href="/blog/topics/internals/">internals<span class="ct">14</span></a>
@@ -125,6 +125,24 @@
</div>
</div>
</article>
+<article class="card">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
+ <span>GUIDES</span>
+ </a>
+ <div class="body">
+ <div class="meta">
+ <time datetime="2026-06-05">Jun 5, 2026</time>
+ <span class="dot" aria-hidden="true"></span>
+ <span>5 min</span>
+ </div>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
+ <div class="foot">
+ <span class="tag kind">Guides</span>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
+ </div>
+ </div>
+</article>
<article class="card">
<a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
diff --git a/docs/blog/topics/index.html b/docs/blog/topics/index.html
index b5feea99..a0e0e953 100644
--- a/docs/blog/topics/index.html
+++ b/docs/blog/topics/index.html
@@ -70,7 +70,7 @@
<span class="thumb t-guides" aria-hidden="true">GUIDES</span>
<div class="body">
<h3 style="color:var(--ink)">guides</h3>
- <p class="dek">20 posts.</p>
+ <p class="dek">21 posts.</p>
<div class="foot"><span class="more">View topic →</span></div>
</div>
</a>
@@ -144,7 +144,7 @@
<div class="filters" style="padding-top:36px">
<span class="eyebrow" style="align-self:center">Tags</span>
- <a class="chip" href="/blog/tags/multi-agent/">Multi-Agent<span class="ct">50</span></a>
+ <a class="chip" href="/blog/tags/multi-agent/">Multi-Agent<span class="ct">51</span></a>
<a class="chip" href="/blog/tags/claude-code/">Claude Code<span class="ct">36</span></a>
@@ -192,6 +192,8 @@
<a class="chip" href="/blog/tags/mcp/">MCP<span class="ct">3</span></a>
+ <a class="chip" href="/blog/tags/cost/">Cost<span class="ct">3</span></a>
+
<a class="chip" href="/blog/tags/agentic-ai/">Agentic AI<span class="ct">3</span></a>
<a class="chip" href="/blog/tags/aeo/">AEO<span class="ct">3</span></a>
@@ -224,7 +226,9 @@
<a class="chip" href="/blog/tags/reliability/">Reliability<span class="ct">2</span></a>
- <a class="chip" href="/blog/tags/cost/">Cost<span class="ct">2</span></a>
+ <a class="chip" href="/blog/tags/model-routing/">Model Routing<span class="ct">2</span></a>
+
+ <a class="chip" href="/blog/tags/prompt-caching/">Prompt Caching<span class="ct">2</span></a>
<a class="chip" href="/blog/tags/protocols/">Protocols<span class="ct">2</span></a>
@@ -254,8 +258,6 @@
<a class="chip" href="/blog/tags/debugging/">Debugging<span class="ct">1</span></a>
- <a class="chip" href="/blog/tags/model-routing/">Model Routing<span class="ct">1</span></a>
-
<a class="chip" href="/blog/tags/evaluation/">Evaluation<span class="ct">1</span></a>
<a class="chip" href="/blog/tags/ai-search/">AI Search<span class="ct">1</span></a>
@@ -268,8 +270,6 @@
<a class="chip" href="/blog/tags/trust/">Trust<span class="ct">1</span></a>
- <a class="chip" href="/blog/tags/prompt-caching/">Prompt Caching<span class="ct">1</span></a>
-
<a class="chip" href="/blog/tags/prompt-injection/">Prompt Injection<span class="ct">1</span></a>
<a class="chip" href="/blog/tags/coding-agents/">Coding Agents<span class="ct">1</span></a>
diff --git a/docs/blog/topics/internals/index.html b/docs/blog/topics/internals/index.html
index 57492949..31eb1a09 100644
--- a/docs/blog/topics/internals/index.html
+++ b/docs/blog/topics/internals/index.html
@@ -68,7 +68,7 @@
<div class="filters" role="navigation" aria-label="Topics">
<a class="chip" href="/blog/">All</a>
- <a class="chip" href="/blog/topics/guides/">guides<span class="ct">20</span></a>
+ <a class="chip" href="/blog/topics/guides/">guides<span class="ct">21</span></a>
<a class="chip" href="/blog/topics/internals/" aria-current="page">internals<span class="ct">14</span></a>
diff --git a/docs/blog/topics/memory/index.html b/docs/blog/topics/memory/index.html
index ea89310f..b9629098 100644
--- a/docs/blog/topics/memory/index.html
+++ b/docs/blog/topics/memory/index.html
@@ -68,7 +68,7 @@
<div class="filters" role="navigation" aria-label="Topics">
<a class="chip" href="/blog/">All</a>
- <a class="chip" href="/blog/topics/guides/">guides<span class="ct">20</span></a>
+ <a class="chip" href="/blog/topics/guides/">guides<span class="ct">21</span></a>
<a class="chip" href="/blog/topics/internals/">internals<span class="ct">14</span></a>
diff --git a/docs/blog/topics/orchestration/index.html b/docs/blog/topics/orchestration/index.html
index 28d4922d..78732a46 100644
--- a/docs/blog/topics/orchestration/index.html
+++ b/docs/blog/topics/orchestration/index.html
@@ -68,7 +68,7 @@
<div class="filters" role="navigation" aria-label="Topics">
<a class="chip" href="/blog/">All</a>
- <a class="chip" href="/blog/topics/guides/">guides<span class="ct">20</span></a>
+ <a class="chip" href="/blog/topics/guides/">guides<span class="ct">21</span></a>
<a class="chip" href="/blog/topics/internals/">internals<span class="ct">14</span></a>
diff --git a/docs/blog/topics/story/index.html b/docs/blog/topics/story/index.html
index 1e8ee3f6..ad3837aa 100644
--- a/docs/blog/topics/story/index.html
+++ b/docs/blog/topics/story/index.html
@@ -68,7 +68,7 @@
<div class="filters" role="navigation" aria-label="Topics">
<a class="chip" href="/blog/">All</a>
- <a class="chip" href="/blog/topics/guides/">guides<span class="ct">20</span></a>
+ <a class="chip" href="/blog/topics/guides/">guides<span class="ct">21</span></a>
<a class="chip" href="/blog/topics/internals/">internals<span class="ct">14</span></a>
diff --git a/docs/blog/topics/use-cases/index.html b/docs/blog/topics/use-cases/index.html
index dab4a15c..9055c85c 100644
--- a/docs/blog/topics/use-cases/index.html
+++ b/docs/blog/topics/use-cases/index.html
@@ -68,7 +68,7 @@
<div class="filters" role="navigation" aria-label="Topics">
<a class="chip" href="/blog/">All</a>
- <a class="chip" href="/blog/topics/guides/">guides<span class="ct">20</span></a>
+ <a class="chip" href="/blog/topics/guides/">guides<span class="ct">21</span></a>
<a class="chip" href="/blog/topics/internals/">internals<span class="ct">14</span></a>
diff --git a/docs/blog/trigger-ai-agents-from-slack/index.html b/docs/blog/trigger-ai-agents-from-slack/index.html
index eda4c356..929bc8c0 100644
--- a/docs/blog/trigger-ai-agents-from-slack/index.html
+++ b/docs/blog/trigger-ai-agents-from-slack/index.html
@@ -304,20 +304,20 @@ it’s free and open source.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/blog/whats-new-in-agentic-ai-june-2026/index.html b/docs/blog/whats-new-in-agentic-ai-june-2026/index.html
index 49541848..c300f01b 100644
--- a/docs/blog/whats-new-in-agentic-ai-june-2026/index.html
+++ b/docs/blog/whats-new-in-agentic-ai-june-2026/index.html
@@ -362,20 +362,20 @@ hive run; it’s free and open source.</p>
</div>
</article>
<article class="card">
- <a class="thumb t-guides" href="/blog/agent-security-and-sandboxing/" aria-hidden="true" tabindex="-1">
+ <a class="thumb t-guides" href="/blog/the-multi-agent-cost-playbook/" aria-hidden="true" tabindex="-1">
<span>GUIDES</span>
</a>
<div class="body">
<div class="meta">
- <time datetime="2026-06-04">Jun 4, 2026</time>
+ <time datetime="2026-06-05">Jun 5, 2026</time>
<span class="dot" aria-hidden="true"></span>
<span>5 min</span>
</div>
- <h3><a href="/blog/agent-security-and-sandboxing/">Running AI Agents Safely: Permission Modes and Sandboxing</a></h3>
- <p class="dek">How to run autonomous AI agents safely: permission modes and the bypassPermissions foot-gun, workspace isolation, and what to gate vs. allow.</p>
+ <h3><a href="/blog/the-multi-agent-cost-playbook/">The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend</a></h3>
+ <p class="dek">Caching, batching, model tiering, context discipline, local-first — the five levers that cut a multi-agent fleet&#39;s bill, and how they compound.</p>
<div class="foot">
<span class="tag kind">Guides</span>
- <a class="more" href="/blog/agent-security-and-sandboxing/" aria-label="Read: Running AI Agents Safely: Permission Modes and Sandboxing">Read →</a>
+ <a class="more" href="/blog/the-multi-agent-cost-playbook/" aria-label="Read: The Multi-Agent Cost Playbook: Five Levers to Cut AI Agent Spend">Read →</a>
</div>
</div>
</article>
diff --git a/docs/sitemap.xml b/docs/sitemap.xml
index 66799d96..827c3191 100644
--- a/docs/sitemap.xml
+++ b/docs/sitemap.xml
@@ -10,7 +10,7 @@
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
<url>
<loc>https://munderdiffl.in/</loc>
- <lastmod>2026-06-04</lastmod>
+ <lastmod>2026-06-05</lastmod>
<changefreq>weekly</changefreq>
<priority>1.0</priority>
</url>
@@ -35,6 +35,12 @@
<changefreq>monthly</changefreq>
<priority>0.8</priority>
</url>
+ <url>
+ <loc>https://munderdiffl.in/blog/the-multi-agent-cost-playbook/</loc>
+ <lastmod>2026-06-05</lastmod>
+ <changefreq>monthly</changefreq>
+ <priority>0.8</priority>
+ </url>
<url>
<loc>https://munderdiffl.in/blog/agent-security-and-sandboxing/</loc>
<lastmod>2026-06-04</lastmod>
@@ -499,6 +505,7 @@
<url><loc>https://munderdiffl.in/blog/tags/terminals/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/electron/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/mcp/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
+ <url><loc>https://munderdiffl.in/blog/tags/cost/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/agentic-ai/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/aeo/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/seo/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
@@ -515,7 +522,8 @@
<url><loc>https://munderdiffl.in/blog/tags/performance/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/architecture/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/reliability/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
- <url><loc>https://munderdiffl.in/blog/tags/cost/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
+ <url><loc>https://munderdiffl.in/blog/tags/model-routing/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
+ <url><loc>https://munderdiffl.in/blog/tags/prompt-caching/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/protocols/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/subagents/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/markdown/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
@@ -530,14 +538,12 @@
<url><loc>https://munderdiffl.in/blog/tags/error-handling/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/context-engineering/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/debugging/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
- <url><loc>https://munderdiffl.in/blog/tags/model-routing/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/evaluation/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/ai-search/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/cloud-agents/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/long-running-agents/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/trends/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/trust/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
- <url><loc>https://munderdiffl.in/blog/tags/prompt-caching/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/prompt-injection/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/coding-agents/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>
<url><loc>https://munderdiffl.in/blog/tags/hitl/</loc><changefreq>weekly</changefreq><priority>0.3</priority></url>