Change Request: llm.txt file to let AI agents follow the latest Eslint documentation accordingly with new new updates

[HatemMn]

ESLint version

latest

What problem do you want to solve?

Despite the lack of official status, a lot of websites have this file ( source )

Frameworks like Astro and Mintlify seem to generate it automatically.
Libraries like Ant Design have comprehensive ones, their documentation is well done : https://ant-design.antgroup.com/docs/react/llms

Two variants have emerged in practice:

• /llms.txt — a curated index with links to key docs
• /llms-full.txt — a single concatenated markdown dump of the entire site

This is just an idea to discuss. Eslint is used in almost every serious project but rare are the devs who still configure it by hand and I really tend to see a lot of configs that are not even migrated to the flat configuration.

If the idea is accepted, I will try to help code it

What do you think is the correct solution?

I am not well versed in the architecture of this repo, but I think it should be auto generated to avoid release friction

• /llms.txt — a curated index with links to key docs
• /llms-full.txt — a single concatenated markdown dump of the entire site

Participation

• I am willing to submit a pull request for this change.

AI acknowledgment

• I did not use AI to generate this issue report.
• (If the above is not checked) I have reviewed the AI-generated content before submitting.

Additional comments

No response

[DMartens]

I agree that an llms.txt may be helpful, as we are able to curate the content (sitemap.xml only offers an index).
I think the key documents are (based on the primary use cases):

• Configuring ESLint (How to ignore files, ...)
• How to resolve a problem reported by a rule
• Creating a rule

[nzakas]

This may be trickier than it seems. I'm assuming the point of llms.txt would be to find documentation, but the docs don't actually live in the eslint.org repo, they live in the eslint repo, so there would be some complexity with autogenerating.

I'm also not sure how useful this is going to be in the future vs something like a docs MCP server.

[HatemMn]

@nzakas for the question

I'm also not sure how useful this is going to be in the future vs something like a docs MCP server.

I understand your point but IMO both serve a different pupose and in an ideal world we would have both

• MCP server is for devs who are aware of the importance of Eslint and are either creating rules ou managing a very complex set of rules that would require an MCP server for their LLM to dig in
• the llm.txt is something I see in the same way as robots.txt files, it's just some static files that aliments crawlers and especially the RAG (exp: perplexity, and most modern tools that search the web)

If an user asks a "web searching" ai to search for docs, it's always better to provide easy access or at least an index for the model than leave it be. They manage to do it without this file, but as I work with antD, I noticed the quality difference in the responses honestly and when I "check sources" I see that my model pulls llm.txt files

Hence the proposal

[nzakas]

That's a fair point.

As I mentioned, there's some technical complexity related to autogenerating llm.txt right now because the docs site is separate from eslint.org even though it's proxied under eslint.org. We could potentially just create a static one (truth be told, the structure of the docs doesn't change all that much) for now and revisit later. That should at least get some better visibility into the docs structure.