At the root of the domain that serves your docs, for example https://help.example.com/llms.txt, served as plain text.
Free tool
llms.txt generator and guide for help centers
Build a valid llms.txt for your docs in a minute, then learn what it does, what it does not do, and where it fits next to sitemap.xml, robots.txt and MCP.
llms.txt is a plain Markdown file served at the root of a site (/llms.txt) that gives large language models a short, curated map of your most useful pages. Per the llmstxt.org proposal it contains an H1 with the site name, a blockquote summary, optional details, and H2 sections listing Markdown links. For a help center, it points AI tools straight to your canonical articles instead of leaving them to guess from navigation and boilerplate.
Free llms.txt generator
Your llms.txt
# Your help center
## Getting started
- [Create your first portal](https://help.example.com/en/articles/create-portal): Set up a branded help center in 5 minutes
- [Invite your team](https://help.example.com/en/articles/invite-team)
## Billing
- [Plans and limits](https://help.example.com/en/articles/plans)
- Add a name: llms.txt must start with an H1 title.
- Add a summary: the blockquote is what most readers use to understand the site.
What llms.txt is
llms.txt is a proposal published at llmstxt.org for a small, human-readable file that helps language models use a website at inference time, for example when an assistant browses your site to answer a question. HTML pages are full of navigation, scripts and layout; an LLM with a limited context window benefits from a concise index that says what the site is and which pages matter.
The format is deliberately simple Markdown, in this order:
- An H1 with the name of the project or site. This is the only required element.
- A blockquote with a short summary that explains what the site covers.
- Optional free-form details: paragraphs or lists with context on how to read the content.
- Zero or more H2 sections, each a list of Markdown links: [Title](URL), optionally followed by a colon and a short note.
- An H2 named "Optional" marks links that can be skipped when context is tight.
Why help centers benefit
- Help centers are exactly the content assistants are asked about: how to set something up, why something fails, what a plan includes.
- A curated list steers models to canonical, current articles instead of outdated forum threads or cached copies.
- Short summaries per link help a model pick the right article for a question without fetching every page.
- It costs almost nothing: one static file, no new infrastructure, easy to keep in version control or regenerate.
- Writing it forces a useful audit: if you cannot list your 20 most important articles, your readers probably cannot find them either.
How to write an llms.txt for your help center
- Pick the name and write a one-sentence summary of what your product does and what the help center covers.
- Add a short details paragraph if there is context a model should know, such as which product versions the docs cover.
- Group your most important articles into a few H2 sections that mirror your categories, for example Getting started, Billing, Troubleshooting.
- For each article add a Markdown link with its canonical URL and, where it helps, a short description after a colon.
- Put nice-to-have links such as the changelog under an "Optional" section.
- Serve the file as plain text at https://your-domain/llms.txt and update it when you add or restructure key articles.
Example llms.txt for a help center
A complete file for a fictional scheduling app. The generator above produces the same structure from your inputs.
# Acme Help Center
> Official help center for Acme, a scheduling app for clinics and salons.
> Covers setup, calendar sync, payments, notifications and billing.
Articles are maintained by the Acme support team and updated with every release.
Prefer the linked pages over older forum posts.
## Getting started
- [Create your account](https://help.acme.com/en/articles/create-account): Sign-up and first login
- [Set up your first calendar](https://help.acme.com/en/articles/first-calendar)
- [Invite team members](https://help.acme.com/en/articles/invite-team): Roles and permissions
## Payments
- [Accept card payments](https://help.acme.com/en/articles/card-payments)
- [Refunds and disputes](https://help.acme.com/en/articles/refunds)
## Billing
- [Plans and limits](https://help.acme.com/en/articles/plans)
- [Change or cancel your plan](https://help.acme.com/en/articles/cancel)
## Optional
- [Changelog](https://help.acme.com/en/changelog): Release notes by month
llms.txt vs sitemap.xml vs robots.txt vs MCP
They complement each other. Keep robots.txt for permissions and sitemap.xml for search engines; add llms.txt as a lightweight map for models; use MCP when you want assistants to query your docs directly instead of reading a static list.
| llms.txt | sitemap.xml | robots.txt | MCP server | |
|---|---|---|---|---|
| Purpose | Curated guide to key pages for LLMs | Complete list of URLs for search engine crawlers | Crawl permissions for bots | Live tools an AI assistant calls to search and read content |
| Format | Markdown | XML | Plain text directives | JSON-RPC over HTTP |
| Coverage | Hand-picked subset | Every indexable page | Rules, not content | Whole published knowledge base, queried on demand |
| Freshness | As fresh as the last edit | Usually generated automatically | Rarely changes | Always current |
| Status | Community proposal | Established standard | Established standard | Open protocol, growing client support |
Limits worth knowing
llms.txt is a proposal, not a W3C or IETF standard, and AI providers have not universally committed to reading it. Treat it as a cheap, low-risk signal rather than a ranking lever. It also goes stale: a hand-written list does not know when you publish, rename or retire an article, so tie updates to your release or content review process.
How Multibase fits
Multibase publishes its own llms.txt at multibase.io/llms.txt. Customer portals do not generate an llms.txt automatically today; use the generator on this page and host the file yourself.
What every Multibase portal does have is a public, read-only MCP endpoint at https://be.multibase.io/api/mcp/<portal-slug>. Instead of a static list, AI assistants such as Claude or Cursor can search your published articles and fetch them as Markdown with canonical URLs, always reflecting what is live. Portals are also server-rendered with a per-portal sitemap, so search engines and crawlers see clean HTML.
Frequently asked questions
There is no evidence that it does. It targets language models reading your site, not search engine ranking. Keep sitemap.xml for search engines.
No. List the pages that answer the most common questions and link to category or index pages for the rest. Put secondary links under an "Optional" section.
A common companion convention: a single file containing the full text of your docs. It suits small doc sets; for large help centers it quickly becomes too big for a model's context.
Not automatically today. Use the free generator above. Every Multibase portal does expose an MCP endpoint that AI assistants can query directly.
Give AI assistants live access to your docs
Every Multibase portal ships with a read-only MCP endpoint. Start on the free plan.