Make AI agents
understand your business
Your website speaks HTML — AI agents don't. MAKO is an open protocol that gives every page a structured, token-efficient version that LLMs actually understand.
The web was designed for humans
When an AI agent visits your site, it downloads navigation bars, cookie banners, ad scripts, and thousands of lines of markup — just to find a product name and price.
average page weight an AI agent must parse
4,125 tokens consumed before understanding a single product
Token Efficiency
Real-world comparison of the same product page across different formats.
What agents get today
~4,125 tokens
With MAKO
~276 tokens
reduction in token consumption
Token counts are model-agnostic estimates (~4 chars/token), a standard approximation across BPE tokenizers.
Who Benefits
MAKO bridges the gap between your content and the AI agents trying to understand it.
Site Owners & Marketers
Your products and content become visible to ChatGPT, Perplexity, and every AI shopping assistant. No code changes — install a plugin and let AI agents find, understand, and recommend your business.
Check your ScoreDevelopers & Platform Teams
A typed SDK, CLI validation, and Express middleware to add MAKO to any stack. Content negotiation, CEF embeddings, and semantic actions — all spec-compliant out of the box.
View on GitHubAI Agent & MCP Developers
Stop wasting tokens parsing HTML. MAKO gives your agent structured frontmatter, declared actions, and semantic links — with a HEAD request that reveals relevance before you download.
Get the SDKWorks With Your Existing Stack
MAKO doesn't replace other standards — it complements them. Each solves a different part of the AI-readiness puzzle.
llms.txt
Site-level instructions for LLMs — what to crawl, what to skip
Page-level structured content with frontmatter, actions, and token-efficient body
Schema.org
Structured metadata for search engines (JSON-LD, microdata)
Full readable content optimized for LLM consumption, not just metadata snippets
WebMCP
Exposing server tools and capabilities to AI agents via MCP
The content itself — products, articles, docs — in a format agents can read directly
Cloudflare /md
HTML-to-Markdown conversion at the edge for any page
Curated, frontmatter-rich documents with 60% fewer tokens than generic markdown
How AI-ready is your site?
Enter your URL and get an instant MAKO Score from 0 to 100.
How It Works
A 6-layer architecture that turns any website into an AI-optimized endpoint — built entirely on standard HTTP.
Discovery
The agent finds which URLs support MAKO via .well-known/mako, link tags, or llms.txt.
Pre-filter
A HEAD request returns metadata headers — token count, type, entity — without downloading the body.
Content
The full MAKO document: structured frontmatter + clean markdown body in ~94% fewer tokens.
Navigation
Semantic links with context let agents navigate your site with purpose, not blind crawling.
Actions
Machine-readable CTAs (buy, subscribe, book) with endpoints and parameters — no JavaScript needed.
Cache
ETags and conditional requests let agents check for changes without re-downloading content.
What a MAKO File Looks Like
Structured YAML frontmatter + optimized Markdown body. Everything an AI agent needs, nothing it doesn't.
---
mako: "1.0"
type: product
entity: "Nike Air Max 90"
tokens: 280
language: en
actions:
- name: add_to_cart
endpoint: /api/cart/add
links:
internal:
- url: /category/running
context: "Browse running shoes"
---
# Nike Air Max 90
Mid-range casual running shoe by Nike.
## Key Facts
- Price: 79.99 EUR (was 149.99, -47%)
- Rating: 4.3/5 (234 reviews)
- In stock, sizes EU 38-46The Ecosystem
Everything you need to start serving MAKO content today.
Start serving AI-optimized content
Choose your path. Open protocol, Apache-2.0 licensed, production-ready.
Developer
Add MAKO to any stack with the TypeScript SDK and CLI tools.
npm install @mako-spec/js