llms.txt and llms-full.txt

Learn what the auto-generated llms.txt and llms-full.txt files are, how agents use them, and how to customize them.

Every time you run great-docs build, two special files are generated alongside your documentation site: llms.txt and llms-full.txt. These files give AI coding agents a compact, text-based view of your package’s API so they can answer questions about your library without scraping HTML pages.

What is llms.txt?

The llms.txt file is a concise index of your package’s public API. It lists every documented class, method, and function with a one-line summary and a link to the full reference page. Here is a shortened example of what the file looks like:

# great_docs

> An easy-to-use documentation site generator for Python packages

## Docs

### API Reference

#### Classes
> Core classes and types

- [GreatDocs](https://example.com/reference/GreatDocs.html): GreatDocs class for creating documentation sites

An agent can read this file to quickly discover what your package offers, then follow the links to get full details on specific items.

What is llms-full.txt?

The llms-full.txt file contains the complete rendered documentation for every public API member: parameter descriptions, return types, examples, and notes. It is much longer than llms.txt but gives an agent everything it needs to work with your API without making additional requests.

This file is especially useful for agents that operate in offline or rate-limited environments, since all the information is available in a single download.

How agents use these files

When an agent has access to your documentation URL (for example, https://example.com/), it can fetch https://example.com/llms.txt to get the index or https://example.com/llms-full.txt for full details. Many agents (including those with the Great Docs skill installed) know to look for these files automatically.

The workflow typically looks like this: the agent reads llms.txt to see what classes and functions are available, identifies the relevant items, and then either follows the links or reads the corresponding sections of llms-full.txt for complete documentation.

Generation is automatic

You do not need to configure anything to get these files. Great Docs generates them as part of the standard build pipeline (step 4 in the build sequence). The files are placed in the great-docs/ output directory and deployed alongside the rest of your site.

The content is derived from the same docstrings and type annotations that produce your HTML reference pages. If a function appears in your API reference, it will appear in both llms.txt and llms-full.txt.

Linking to the files

You might want to mention these files in your project README or in a “For AI Agents” section of your documentation. The URLs follow a predictable pattern:

https://<your-site>/llms.txt
https://<your-site>/llms-full.txt

Some AI agent frameworks also look for these files at well-known paths, so simply having them in your site root is often enough for automatic discovery.

Keeping them accurate

Since the files are regenerated on every build, they stay in sync with your actual API. If you add a new function or update a docstring, the next build will reflect those changes in both files. There is nothing extra to maintain.

If you notice that a function is missing from the output, check whether it is excluded via the exclude list in great-docs.yml or whether its name starts with an underscore (private members are excluded by default). Adjusting the exclude configuration or adding the item to the public API will make it appear on the next build.