# AI-native knowledge base

## A knowledge base agents can actually read.

Publish the help center your customers see, plus the files and tools agents need to discover, read, and index the same content.

Three ways into the same published content

1. `/llms.txt` Discover the relevant guide.
2. `/quickstart.md` Read the guide as clean text.
3. `Hosted MCP server` Look up an answer from a connected agent.

## Give agents more than a web page.

`/llms.txt`

### Find the right page

Publish a hosted index that points agents to relevant guides, categories, and support articles.

`/quickstart.md`

### Read the page as Markdown

Give agents a Markdown version of each published content page, with the headings and links they need and no browser rendering step.

`Hosted MCP server`

### Search from an agent’s tools

Connect a compatible agent to a hosted MCP server so it can search or retrieve help-center content during a task.

## One source

## Update the article, not a second knowledge base.

Publish human-readable docs and agent-facing formats from the same content.

## Does this guarantee AI indexing or citations?

No. These formats make your content available to agents that can access and use them. Each tool decides what it fetches, indexes, or cites.

## Is this the same as adding a support chatbot?

No. These outputs give external agents ways to read your documentation. A customer-facing AI answer widget is a separate setup.

## Publish your first agent-readable guide.

Start with an article your customers already need, then inspect the Markdown output.