> For the complete documentation index, see [llms.txt](https://docs.n8n.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.n8n.io/contribute/contribution-guide-for-n8n-docs/agent-friendly-docs.md).

# Agent-friendly docs

What n8n Docs means by agent-friendly documentation, and how the standard relates to the style guide.

People and AI agents both read n8n Docs. What each user type needs from documentation overlaps more than it diverges. However, agents have some particular needs for effectively parsing and navigating the documentation.

This page sets out what agent-friendly means to n8n Docs, and why it matters. These guidelines are useful to contributors writing for the docs, or any user interested in n8n Docs' approach.

## Why this matters for n8n Docs

Two overlapping audiences use n8n Docs through an agent instead of a browser:

* **People ask an AI agent instead of searching.** A growing share of documentation traffic comes from AI assistants and search agents retrieving pages to answer a question, not from a person clicking through the site's navigation to look at a page directly.
* **People build with n8n using a coding agent.** Someone writing a workflow, a node, or an integration increasingly does so with AI coding tools. These tools retrieve n8n Docs directly to ground their work in current, accurate information instead of outdated training data.

## What "agent-friendly" means

The standard breaks down into four layers.

### Machine-readable delivery

Agents need a way to fetch content without rendering a full web page. n8n Docs provides several routes in:

* **Markdown:** Every page is available as clean Markdown, not just rendered HTML.
* **llms.txt:** [`llms.txt`](https://docs.n8n.io/llms.txt) and [`llms-full.txt`](https://docs.n8n.io/llms-full.txt) list every published page for agents that support the format.
* **MCP server:** An MCP server exposes docs directly to MCP clients. See [Connect to the n8n docs MCP server](https://docs.n8n.io/connect/connect-to-n8n-docs-mcp-server).
* **AI search:** AI-powered search (Kapa.ai) answers questions directly from docs, the blog, and the community forum.
* **Context7:** [Context7](https://context7.com/) indexes n8n Docs so in-editor coding agents can retrieve current docs while someone builds.

### Site-wide consistency

* **Terminology:** n8n Docs uses the same term for the same concept everywhere. This reduces confusion for both human users and agents. See [Terminology and naming](/contribute/contribution-guide-for-n8n-docs/terminology.md).
* **Stable URLs:** URLs stay stable across a page rename, with a redirect in place whenever a URL does change.
* **Feature availability:** A fact that depends on an n8n version or plan is placed at the scope it applies to: a hint for a whole page or section, an inline mention in passing, or a table cell for a table row. This means an agent that retrieves only a fragment of a page still sees the correct context. See [Feature availability](/contribute/contribution-guide-for-n8n-docs/style-guide-for-n8n-docs.md#feature-availability).
* **Review process:** New pages and modifications are reviewed for consistency and technical accuracy before they're published, on n8n Docs' [open-access GitHub repository](https://github.com/n8n-io/n8n-docs), by both human reviewers and n8n Docs' AI-reviewer [cubic](https://www.cubic.dev/).

### Structure and navigation

* **Page length:** n8n Docs splits content into focused pages, short enough for an agent to retrieve without truncation. Search and the docs assistant chunk content on `##`/`###` headings, retrieving one section at a time. A page or section that's too small gets merged with unrelated neighbours instead. Over-splitting a topic across small pages measurably lowers answer quality. See [Page length and granularity](/contribute/contribution-guide-for-n8n-docs/style-guide-for-n8n-docs.md#page-length-and-granularity).
* **Self-contained sections:** Each section stands on its own, without relying on information "as mentioned above". Agents may retrieve only a chunk or section of a page, not the full content, sometimes just a heading. See [Keep each section self-contained](/contribute/contribution-guide-for-n8n-docs/style-guide-for-n8n-docs.md#keep-each-section-self-contained).
* **Answer-first opening:** Each page opens with a one-sentence plain-language answer directly after the H1. AI assistants and search surfaces often show just this line as the citation, so a concrete opening increases both citation rate and answer accuracy. See [Lead with a one-sentence answer](/contribute/contribution-guide-for-n8n-docs/style-guide-for-n8n-docs.md#lead-with-a-one-sentence-answer).
* **Cross-linking:** Pages link explicitly to prerequisites, next steps, and related topics, so an agent can follow a path instead of guessing a URL. Interlinked pages also form a topic cluster that AI search reads as a signal of depth and cites far more than standalone pages. See [Link to related pages, prerequisites, and next steps](/contribute/contribution-guide-for-n8n-docs/style-guide-for-n8n-docs.md#link-to-related-pages-prerequisites-and-next-steps).

### Content and examples

* **Text over images:** Instructions live in easily-parseable text, not images. Agents and screen readers receive only an image's alt text and file path, not the picture itself. Screenshots can only confirm what the reader already knows from the words on the page. They never carry meaning alone. See [Images](/contribute/contribution-guide-for-n8n-docs/style-guide-for-n8n-docs.md#images).
* **Worked examples:** n8n Docs strives to provide worked examples for features with code, expression, or configuration surfaces. Cover the common case and cases that break, because agents copy examples directly and can't infer what isn't shown. See [Show worked examples for each feature](/contribute/contribution-guide-for-n8n-docs/style-guide-for-n8n-docs.md#show-worked-examples-for-each-feature).
* **Tabbed content:** Tabbed content stays sparse and short, since a person sees one tab but an agent reads every one. See [Tabbed content](/contribute/contribution-guide-for-n8n-docs/style-guide-for-n8n-docs.md#tabbed-content).

## How this relates to the style guide

This page explains the reasoning. The [contribution guide](/contribute/contribution-guide-for-n8n-docs.md) covers the contribution process, and the [style guide](/contribute/contribution-guide-for-n8n-docs/style-guide-for-n8n-docs.md) has the specific rules to follow. Several of those rules, including page length, self-contained sections, internal linking, feature availability, images, and tabbed content, exist specifically because they make docs work for agents as well as people. If you're contributing, follow the style guide directly. Come back to this page when you want the reasoning behind a rule, or when you're weighing whether something new is worth adding as a rule at all.

## Where this is still evolving

Not every guideline has a settled answer yet. n8n Docs is continuously testing and researching what makes agent-friendly documentation, and developing new systems to help meet these standards. Agent-friendliness is a fast-moving area of documentation practice, and this page will change as the practice matures.

## Related resources

* [Contribution guide for n8n Docs](/contribute/contribution-guide-for-n8n-docs.md)
* [Style guide for n8n Docs](/contribute/contribution-guide-for-n8n-docs/style-guide-for-n8n-docs.md)
* [Terminology and naming](/contribute/contribution-guide-for-n8n-docs/terminology.md)