Transaction-ready context for AI agents
Like llms.txt, but for e-commerce
TL;DR: llms.txt is for content discovery. CommerceTXT is for commerce transactions. If AI needs to know your price, inventory, or shipping policy—you need CommerceTXT.
Key Features
CommerceTXT includes powerful directives for transaction-ready commerce:
@INVENTORY
Real-time stock levels eliminate "Is it available?" hallucinations.
Stock: 42
LowStockThreshold: 10
RestockDate: 2025-12-20Impact: AI can say "Only 3 left" instead of guessing.
@SUBSCRIPTION
Handle recurring payments for SaaS and subscription products.
Plans:
- Monthly: 29.99
- Annual: 299.00
Trial: 7 Days FreeImpact: AI compares one-time vs. subscription pricing accurately.
@REVIEWS
Aggregated review data in 50 tokens instead of scraping 15,000.
Rating: 4.7 / 5.0
Count: 1243
TopTags: "Great battery"Impact: 99.7% token reduction for review queries.
CommerceTXT vs llms.txt
Both protocols help AI understand websites, but they solve different problems:
llms.txt
Use case: Content & Documentation
- 📄 Links to blog posts and docs
- 🔍 Discovery focused
- 💬 "What does this company do?"
- 📊 Static content
CommerceTXT
Use case: E-commerce & Transactions
- 💰 Live pricing and inventory
- 🛒 Transaction ready
- 🤖 "Can I buy this? How much? When ships?"
- ⚡ Dynamic commerce data
Detailed Comparison
| Feature | llms.txt | CommerceTXT |
|---|---|---|
| Primary Use Case | Content discovery | Commerce transactions |
| Data Type | Static URLs | Dynamic pricing, inventory, policies |
| Structure | Flat list | Hierarchical (Root → Category → Product) |
| Updates Needed | When new content published | When inventory/prices change |
| Schema.org Mapping | ❌ No | ✅ Yes (legal compliance) |
| Multi-regional Support | ❌ No | ✅ Built-in (@LOCALES) |
| Real-time Inventory | ❌ No | ✅ @INVENTORY directive |
| Subscription Pricing | ❌ No | ✅ @SUBSCRIPTION directive |
Why Merchants Need CommerceTXT
🎯 Transaction-Ready
llms.txt is for content. CommerceTXT is for commerce. Price, inventory, shipping, returns—everything AI needs to help customers buy.
⚖️ Legal Compliance
Maps to Schema.org. When AI quotes your price or return policy, it's pulling from the same structured data that Google uses. Defensible, verifiable.
🌍 Built for Global Commerce
Not just multi-language—multi-currency, multi-shipping, multi-policy. German customers see EUR prices and GDPR-compliant terms.
💰 ROI is Clear
Every AI shopping query that gets the right answer = potential sale. Wrong price/availability = lost customer. CommerceTXT ensures accuracy.
Real-World Impact
- AI Visibility: Your products appear correctly in ChatGPT, Claude, and Gemini shopping responses
- Reduced Errors: 95% fewer hallucinated prices and policies with @INVENTORY and @REVIEWS
- First-Mover Advantage: Be discoverable in AI commerce before your competitors
- Control Your Narrative: Guide how AI presents your products, policies, and brand voice
Efficiency & Ecology
The current web is built for humans, not agents. Rendering a typical e-commerce page to extract a single price requires massive compute resources.
| Metric | Standard HTML Scraping | CommerceTXT | Impact |
|---|---|---|---|
| Payload Size | ~2.5 MB / page | ~0.005 MB / file | 500x Smaller |
| Token Count (Product) | ~8,500 tokens | ~380 tokens | 95% Reduction |
| Token Count (Reviews) | ~15,000 tokens | ~50 tokens | 99.7% Reduction |
| Compute Cost | High (JS execution) | Zero (Static Text) | CPU Saved |
| Data Integrity | Probabilistic (Hallucinations) | Deterministic (Ground Truth) | Transaction Ready |
🌱 Sustainable AI
Every token processed by an LLM consumes energy. By stripping away the HTML presentation layer and serving pure context, CommerceTXT can reduce the carbon footprint of AI commerce crawling by over 99% (based on internal payload benchmarks). Adopting this standard is a direct contribution to Scope 3 emissions reduction for tech platforms.
How It Works
The protocol uses a Fractal Architecture to minimize token usage. Agents start at the Root and traverse deeper only when user intent requires it.
Level 1: The Root (Store Identity)
Contains store identity, locales, payments, and shipping rules.
# commerce.txt
Version: 1.0
# @IDENTITY
Name: Demo Store
Currency: USD
# @LOCALES
en-US: /commerce.txt (Current)
de-DE: /de/commerce.txt
# @PAYMENT
Methods: Visa, MasterCard, PayPal, ApplePay
# @SHIPPING
- Standard: Free over $50 | 3-5 Days
- Express: $15.00 | 1-2 Days
# @POLICIES
Returns: 30 Days
# @SUPPORT
Email: [email protected]
Chat: Available 9am-5pm EST
# @CATALOG
- Electronics: /categories/electronics.txtLevel 2: The Category (Filters & Promos)
# electronics.txt
# @FILTERS
Brands: Sony, Samsung
Type: Wireless
# @PROMOS
- SAVE10: 10% Off (Expires 2025-12-31)
# @ITEMS
- Sony XM5: /products/xm5.txtLevel 3: The Product (Transaction Data)
# xm5.txt
# @PRODUCT
GTIN: 027242919412
SKU: WH1000XM5
# @OFFER
Price: 348.00
Availability: InStock
# @INVENTORY
Stock: 42
LowStockThreshold: 10
StockStatus: InStock
LastUpdated: 2025-12-14T15:30:00Z
# @SPECS
Battery: 30 Hours
ANC: True
# @IN_THE_BOX
- Sony WH-1000XM5 Headphones
- USB-C charging cable (1.2m)
- Carrying case
Note: AC adapter NOT included
# @REVIEWS
Rating: 4.7
RatingScale: 5.0
Count: 1243
TopTags: "Great battery", "Comfortable"
# @COMPATIBILITY
WorksWith: iPhone 15, iPad ProSubscription Products
# @PRODUCT
Name: Premium Membership
# @SUBSCRIPTION
Plans:
- Monthly: 29.99 | BilledAs: "$29.99/month"
- Annual: 299.00 | BilledAs: "$299/year" | Savings: "Save 17%"
Trial: 7 Days Free
CancelAnytime: TrueProduct Variants (Optional)
# @VARIANTS
Type: Storage
Options:
- 128GB: 999.00 | SKU: GA05843-128
- 256GB: 1099.00 | SKU: GA05843-256
- 512GB: 1299.00 | SKU: GA05843-512
Type: Color
Options:
- Obsidian: +0.00
- Bay: +50.00 | Note: "Limited Edition"For the formal specification, refer to the Full RFC Spec (v1.0).
Future Vision
The following features are proposed for future versions but require community discussion before inclusion:
@ACTIONS EXPERIMENTAL - v2.0+
Vision: Enable AI agents to perform transactional actions—but ONLY with explicit user confirmation (user-in-the-loop).
Example concept:
# @ACTIONS
CheckInventory: GET https://api.example.com/inventory/{sku}
# Read-only, no auth needed
AddToCart: POST https://api.example.com/cart/add
# Requires OAuth2 + user confirmation
UserConfirmation: Required before execution⚠️ CRITICAL: Safety Requirements
- NO autonomous purchases: AI cannot complete transactions without explicit user approval
- Payment isolation: AI MUST NEVER access payment credentials
- Full transparency: Users review cart, price, shipping before authorizing
- Fraud prevention: Rate limiting, audit trails, abuse detection
Why discussion is needed:
- Security frameworks: CSRF protection, OAuth standards, session management
- User consent mechanisms: How to ensure informed, explicit approval
- Liability frameworks: Who is responsible for AI transaction errors?
- Fraud prevention: Preventing malicious agents from abusing APIs
- Should CommerceTXT remain read-only in v1.0? (Recommended: YES)
Recommendation: CommerceTXT v1.0 should remain read-only. Transactional capabilities only after Trust Score system is proven and security standards are established.
Join the discussion: GitHub Discussions
@SUSTAINABILITY RFC Needed
Vision: Verified environmental and ethical claims to reduce greenwashing and help consumers make informed choices.
Example concept:
# @SUSTAINABILITY
CarbonNeutral: Yes
Verified: https://climatepartner.com/cert/12345
Certificate: ClimatePartner ID 12345-2024
Packaging: 100% Recycled Materials
Certification: FSC-C123456
RepairProgram: Available
Duration: 7 years of spare partsWhy discussion is needed:
- Verification requirements: Should third-party proof be mandatory?
- Which certification bodies are globally recognized?
- How to handle regional differences in sustainability standards?
- Anti-greenwashing policies: flagging unverified claims
Join the discussion: GitHub Discussions
Trust & Verification
CommerceTXT minimizes hallucinations by design, but trust must be earned. The protocol includes a verified trust model.
🛡️ Cross-Verification
AI agents periodically verify commerce.txt data against visible HTML and Schema.org. Discrepancies >5% trigger Trust Score penalties.
⏱️ Freshness Check
Agents validate @INVENTORY timestamps. Data older than 24h is flagged; older than 72h is treated as "Unknown".
✅ Verified Reviews
Review data is trusted only if backed by a verifiable Source URL (e.g. Trustpilot, Google Reviews).
Quick Start
For Merchants
Add a single text file to your root directory.
curl -O https://raw.githubusercontent.com/commercetxt/commercetxt/main/templates/commerce.txtEdit with your details and upload to yourdomain.com/commerce.txt
For Developers
Parse deterministic data, not HTML soup.
# Python Example
import requests
# ... parser logic ...
print(data['@OFFER']['Price'])