Use the docs with an AI assistant

These docs are written for people, but they are just as useful to an AI assistant. Every page exists in a plain-text form an assistant can read directly, and the whole site can be pasted into one conversation. Whether you use ChatGPT, Claude or another assistant, the steps are the same.

Copy a page

Every guide has a Copy page menu at the top, above the text. It offers four actions:

• Copy as Markdown copies the page to your clipboard as plain Markdown, ready to paste into a chat.
• View as Markdown opens that plain-text version in a new tab, so you can read or save it.
• Open in ChatGPT and Open in Claude start a new conversation in that assistant with the page's address already in the prompt, so it reads the guide before you ask anything.

Every page has a Markdown address

The plain-text version of a page lives next to the page itself: take the page's address and add .md at the end. For example, the Datasets guide is at /docs/analytical-apps/pipeline/datasets, and its Markdown version is at /docs/analytical-apps/pipeline/datasets.md. The home page is the one exception: it has no page name of its own to add .md to, so its Markdown version is at /docs/index.md, and in German at /docs/de/index.md. An assistant that can read web addresses can be pointed at that file directly, and it will see the same text you see, without the menus around it.

Give an assistant the whole site

Two addresses cover all of the docs at once:

• /docs/llms.txt is the index: every page with its title and address, as a short list. Paste it into a conversation so the assistant knows what exists and can ask for the pages it needs.
• /docs/llms-full.txt is every page in one file. Paste it, or point the assistant at it, when you want it to answer from the complete documentation rather than from a single guide.

What to paste, and what to ask

Start with the page that matches your question. If you are setting up an import, copy the Datasets guide; if you are writing a transform, copy the SQL nodes guide. Paste it, then ask in your own words: "Which column types can I use?" or "Why would my import be rejected?". For a question that spans the product, use the index or the full file instead, and name the feature you are working on so the assistant can find the right section.

Not sure which guide you need? The docs search finds any page by a word it contains; see Search.