# Logitech Digital Commerce Agent Protocols (AGENTS.md)
Version: 1.1.0
Last Updated: 2026-10-30
Canonical Authority: https://www.logitech.com/AGENTS.md

## 1. Overview & Policy Statement
This document defines the operational rules, rate limits, and discovery standards for autonomous AI agents, web crawlers, LLM assistants, and automated procurement bots interacting with `logitech.com`.

Logitech welcomes responsible automated interactions that aid users in discovering product information, pricing, compatibility, and availability. Agents interacting with `logitech.com` are expected to adhere strictly to the protocols detailed herein.

## 2. Core Agent Principles
* **Efficiency First:** Utilize structured JSON-LD schemas, `llms.txt` mirrors, and dedicated API endpoints over raw HTML scraping whenever available.
* **Identify Transparently:** Provide a distinct `User-Agent` header containing your bot name and contact URL/email.
* **Respect Limits:** Honor HTTP `429` status codes and `Retry-After` headers immediately.
* **Protocol Precedence:** In the event of a conflict between directives in `robots.txt` and `AGENTS.md`, the `robots.txt` disallow rules take precedence.
* **Authentication & Elevated Access:** Agents requiring higher rate limits or access to non-public commerce endpoints must provide an authorized API key or OAuth 2.0 bearer token in the `Authorization` header.

## 3. Product Detail Page (PDP) Interaction Guidance
### 3.1 Primary Canonical Reference Example
For all product-level operations, agents should use canonical URL patterns. The following primary URL illustrates standard interaction mechanics:

* **Canonical PDP URL:** `https://www.logitech.com/en-us/shop/p/mx-master-4`
* **Product Name:** MX Master 4 Wireless Mouse
* **Supported Locale Structure:** `https://www.logitech.com/{language}-{country}/shop/p/{slug}`

### 3.2 Fetching Strategy
When evaluating a PDP such as `https://www.logitech.com/en-us/shop/p/mx-master-4`:

1. **Do First** check `https://www.logitech.com/llms.txt` for direct markdown mirrors or attempt fetching markdown versions by appending `.md` (e.g., `https://www.logitech.com/en-us/shop/p/mx-master-4.md`).
2. **If Unavailable**, fetch the static HTML response and extract the embedded Schema.org JSON-LD payload located inside `<script type="application/ld+json">`.
3. **Do Not** execute client-side JavaScript rendering engines (e.g., Puppeteer, Playwright) if structured data or markdown alternatives are available.

### 3.3 Sample Expected JSON-LD Structure
Agents targeting `https://www.logitech.com/en-us/shop/p/mx-master-4` should expect data formatted according to Schema.org standard:

```json
{
  "@context": "https://schema.org/",
  "@type": "Product",
  "name": "MX Master 4 Wireless Mouse",
  "brand": {
    "@type": "Brand",
    "name": "Logitech"
  },
  "offers": {
    "@type": "Offer",
    "url": "https://www.logitech.com/en-us/shop/p/mx-master-4",
    "priceCurrency": "USD",
    "price": "119.99",
    "itemCondition": "https://schema.org/NewCondition",
    "availability": "https://schema.org/InStock"
  }
}
```

## 4. Rate Limits & Traffic Management
Agents must strictly comply with the rate limits defined below:
* **HTML PDP Pages:** Max 30 req/min (Concurrency: 2 connection limit, Exponential backoff on 429)
* **Headless JSON APIs:** Max 120 req/min (Concurrency: 5 connection limit, Obey `Retry-After` header)
* **Search & Auto-suggest:** Max 20 req/min (Concurrency: 1 connection limit, Fixed 5-second delay)

## 5. Alternate Endpoints & Machine-Readable Discovery
To reduce load on web infrastructure, agents should utilize structured discovery protocols:

### 5.1 LLM Content Index (llms.txt)
For curated markdown product documentation and indexes designed specifically for LLM ingestion, visit:
`https://www.logitech.com/llms.txt`

### 5.2 Product Sitemaps
For comprehensive URL discovery and catalog crawling, refer to the master sitemap index:
`https://www.logitech.com/sitemap.xml`

## 6. Allowed & Disallowed Actions
### Allowed
* Parsing pricing, stock availability, and technical specifications for search indexing, comparison shopping, or consumer assistance.
* Caching product structured data up to 24 hours (86,400 seconds).

### Disallowed
* Automated checkout execution or cart reservation without explicit user authentication and OAuth authorization.
* Bypassing rate-limiting mechanisms via proxy rotation or header spoofing.
* Scraping customer review text for model training without authorization.

## 7. Contact & Security Response
Refer to `https://www.logitech.com/.well-known/security.txt` for security contact information, vulnerability disclosure guidelines, and bot governance policies.
