Alibaba Cloud Docs Are Now Fully Available as llms.txt — 300+ Cloud Products, One URL for Every AI Agent

xianzhong li — Thu, 16 Jul 2026 08:19:30 +0000

TL;DR — All Alibaba Cloud product docs, API references and best-practice guides are now indexed as llms.txt. Whether you use Qwen, GPT, Claude, DeepSeek, Kimi or Doubao, one URL is enough to plug the entire Alibaba Cloud knowledge base into your model or coding agent.

1. Where to find llms.txt

Alibaba Cloud docs live on two independent sites, together shipping six language editions of llms.txt:

International site (alibabacloud.com)

Language	llms.txt
English	https://www.alibabacloud.com/help/en/llms.txt
中文	https://www.alibabacloud.com/help/zh/llms.txt
日本語	https://www.alibabacloud.com/help/ja/llms.txt
Bahasa Indonesia	https://www.alibabacloud.com/help/id/llms.txt

China site (help.aliyun.com)

Language	llms.txt
中文	https://help.aliyun.com/zh/llms.txt
English	https://help.aliyun.com/en/llms.txt

Root indexes

The multi-language entries are aggregated at:

International root: https://www.alibabacloud.com/help/llms.txt
China root: https://help.aliyun.com/llms.txt

2. What llms.txt looks like

Every file follows the llmstxt.org convention — a small Markdown document that tells an AI what a site contains, what each page is about, and whether it's worth fetching:

# Product name
> One-line product summary (so the AI can decide relevance at a glance)

## Section
- [Page title](URL): One-line page summary (what problem this page solves)
- [Page title](URL): ...

Alibaba Cloud uses a two-level index:

Level 1 — Root index (alibabacloud.com/help/en/llms.txt): lists every product; each entry points to that product's own llms.txt.

# Alibaba Cloud Documentation
> Alibaba Cloud is one of the world's leading cloud computing and AI companies...

## AI & Machine Learning
- [PAI](https://www.alibabacloud.com/help/en/pai/llms.txt): End-to-end AI platform — DSW modeling, DLC distributed training, EAS inference...
- [Model Studio](https://www.alibabacloud.com/help/en/model-studio/llms.txt): OpenAI-compatible API for Qwen, DeepSeek and more...

## Storage
- [OSS](https://www.alibabacloud.com/help/en/oss/llms.txt): Object storage service for any type and any scale of data...

## Compute
- [ECS](https://www.alibabacloud.com/help/en/ecs/llms.txt): Elastic compute for on-demand VM provisioning and management...
... 300+ products in total

Level 2 — Product index (e.g. alibabacloud.com/help/en/oss/llms.txt): lists every documentation page for that product.

# Object Storage Service (OSS)
> OSS is an IaaS storage service for storing and accessing any type of data on demand...

## User Guide
- [What is OSS](https://www.alibabacloud.com/help/en/oss/user-guide/what-is-oss.md): Core concepts (Bucket/Object/Endpoint/AccessKey), how OSS works...
- [Configure CORS](https://www.alibabacloud.com/help/en/oss/user-guide/configure-cross-origin-resource-sharing.md): Configure CORS rules to fix browser cross-origin errors...
- [Console quick start](https://www.alibabacloud.com/help/en/oss/user-guide/console-quick-start.md): Create a Bucket and upload/download an Object in 10 minutes...

## API Reference
- [PutBucketCors](https://www.alibabacloud.com/help/en/oss/developer-reference/putbucketcors.md): API reference for setting CORS rules...
...

The two levels link differently:

Root links point to product llms.txt files (e.g. /en/oss/llms.txt).
Product links point to actual doc pages (e.g. /en/oss/user-guide/xxx.md).

An agent reads the root index to locate the product, reads the product index to locate the page, and only then fetches the Markdown body — three hops, load on demand, no wasted tokens.

3. How AI models consume llms.txt

User question
  "How do I deploy a GPU instance on ECS?"
      │
      ▼
┌─────────────────────────────┐
│ Step 1: fetch root index    │  GET alibabacloud.com/help/en/llms.txt
│ → locate ECS entry          │  → /help/en/ecs/llms.txt
└─────────────┬───────────────┘
              ▼
┌─────────────────────────────┐
│ Step 2: fetch product index │  GET /help/en/ecs/llms.txt
│ → match relevant pages      │  → "GPU instance families", "Create a GPU instance"
└─────────────┬───────────────┘
              ▼
┌─────────────────────────────┐
│ Step 3: fetch target page   │  GET the concrete doc URL (.md or HTML)
│ → answer the user           │
└─────────────────────────────┘

This lets the model navigate straight to the right doc instead of guessing across tens of thousands of pages.

4. Usage examples

Example 1 — Use it directly inside AI chat products

Chat products with built-in web-fetch tools (Qoder, Kimi, ChatGPT, Claude, Perplexity, Doubao, and so on) can consume llms.txt from a plain-language instruction. No code required.

Just ask the model:

Read https://www.alibabacloud.com/help/en/llms.txt ,
find the Function Compute product, and show me how to
write an HTTP-triggered function in Python.

What the model actually does:

1. Fetch(https://www.alibabacloud.com/help/en/llms.txt)                       ← root index
   → find the Function Compute entry, get the product llms.txt URL

2. Fetch(https://www.alibabacloud.com/help/en/functioncompute/llms.txt)       ← product index
   → keyword-match HTTP trigger / Python runtime pages

3. Fetch(https://www.alibabacloud.com/help/en/functioncompute/.../http-trigger-invoking-function.md)  ← page
   → read the body and produce the answer

The model performs the "root → product → page" hop chain automatically, grounds the answer in real docs, and cites the source URL.

Tip — cited URLs are .md by default (AI-friendly). Strip .md to get the human-readable HTML page:

AI returns: https://www.alibabacloud.com/help/en/functioncompute/fc/user-guide/http-trigger-invoking-function.md

Open in browser: https://www.alibabacloud.com/help/en/functioncompute/fc/user-guide/http-trigger-invoking-function

Wire it in as a knowledge source in your System Prompt (works for custom GPTs, Kimi agents, Qoder personas, etc.):

You are an Alibaba Cloud architect assistant.
The index of every Alibaba Cloud product doc lives at
https://www.alibabacloud.com/help/en/llms.txt
When answering, first locate the relevant product via the index,
then fetch the specific page. When you cite a URL, strip the trailing
.md so the user opens the human-readable HTML.

Example 2 — Consume llms.txt from Qoder CLI (Tongyi Lingma CLI)

Tongyi Lingma is Alibaba Cloud's AI coding assistant. Its CLI mode (qodercli) ships a built-in WebFetch tool, so you can teach it to navigate llms.txt by dropping a project-level AGENTS.md file.

Set-up. Create AGENTS.md in your project root:

# Alibaba Cloud docs lookup rules

Whenever you need Alibaba Cloud product documentation, follow these steps:

1. Fetch https://www.alibabacloud.com/help/en/llms.txt to get the product index.
2. Locate the product's own llms.txt (e.g. ECS → /help/en/ecs/llms.txt).
3. From the product index, pick the most relevant page.
4. Fetch that page for the full content.
5. When citing a URL, strip the trailing `.md` so the user gets the HTML page.

Result (verified):

$ qodercli

> How do I configure CORS on OSS? Please follow AGENTS.md,
> use Alibaba Cloud llms.txt to find the OSS docs, then answer.

Lingma automatically:

Fetches https://www.alibabacloud.com/help/en/llms.txt → finds the OSS entry.
Fetches https://www.alibabacloud.com/help/en/oss/llms.txt → finds the CORS page.
Grounds the answer in the real page and cites https://www.alibabacloud.com/help/en/oss/user-guide/configure-cross-origin-resource-sharing.md.

Why it works — Qoder CLI loads three levels of AGENTS.md (user-level ~/.qoder/AGENTS.md, project-level, workspace-local) as context. Combined with WebFetch, the model follows the rules to walk llms.txt on its own. The same pattern works with Claude Code (CLAUDE.md) and any other agentic CLI that supports project-level system prompts.

5. Product coverage

llms.txt covers 300+ Alibaba Cloud products. Major categories include:

Category	Representative products
AI & Machine Learning	PAI, Model Studio, DashVector, AIRec, etc.
Compute	ECS, GPU ECS, Simple Application Server, ECI, etc.
Container	Container Service for Kubernetes (ACK), Container Compute Service, Service Mesh, ECI, etc.
Storage	OSS, NAS, CPFS, HDFS, etc.
Networking & CDN	SLB, EIP, Anycast EIP, CDT, etc.
Security	Anti-DDoS, WAF, Cloud Firewall, Security Center, etc.
Middleware	MSE, EDAS, AHAS, Service Mesh, etc.
Database	PolarDB, RDS, Tair (Redis-compatible), MongoDB, etc.
Big-Data Compute	MaxCompute, Hologres, Realtime Compute for Apache Flink, Elasticsearch, etc.
Media Services	ApsaraVideo Live, ApsaraVideo VOD, RTC, IMS, etc.
Enterprise Services & Cloud Communications	SMS, VMS, Number Privacy Protection, DirectMail, etc.
Domains & Websites	Domains, ICP Filing, Alibaba Cloud DNS, Trademark Service, etc.
End-User Computing	Wuying WorkSpace, Wuying Personal, Wuying AppStreaming, Wuying Terminal, etc.
IoT	IoT Wireless Connectivity, IoT Platform, IoT Device Identity Auth, etc.
Developer Tools	Cloud Shell, OpenAPI Explorer, Alibaba Cloud CLI, Tongyi Lingma, etc.
Serverless	Function Compute, SAE, CAP, CloudFlow, etc.
Migration & O&M	SLS, ARMS, Managed Service for Prometheus, Managed Service for Grafana, etc.
Solutions	DingTalk Dedicated Edition, Government Cloud, SAP on Alibaba Cloud, etc.
Support & Services	Management Console, Account Center, Ticket System API, Support & Services, etc.

6. FAQ

Q: How is llms.txt different from traditional docs search?
Search engines require the user to craft the right keywords, and models struggle to navigate tens of thousands of pages on their own. llms.txt gives models a structured table of contents with descriptions, so they can browse it the way a developer reads an API index — cutting token cost and hallucination rate.

Q: How often is llms.txt updated?
Alibaba Cloud's llms.txt refreshes with the docs themselves. Whenever a product publishes or updates a page, the corresponding llms.txt entry and the .md file update accordingly.

Q: I'm a developer. How do I wire this into my own AI app?
Add the llms.txt URL into your app's System Prompt and tell the model to use it as the canonical Alibaba Cloud knowledge source. Models do not discover llms.txt on their own — you must reference the URL explicitly.

Q: How do the China site and the international site differ?
The two sites cover different product catalogs and, in some cases, different features per product. The llms.txt indexes are maintained independently; language and content don't cross over. Pick the site that matches your deployment region.

DEV Community: xianzhong li