What is llms-full.txt?
llms-full.txt is the companion file to /llms.txt — same llmstxt.org convention, but instead of listing your important pages, it contains the entire concatenated text content of your site in a single file. One fetch, the whole corpus.
When does it help?
llms.txt is a directory. An agent reads it, picks a relevant URL, fetches that page, repeats. That works well for agents that can browse.
llms-full.txt is the full text. An agent reads it once and has everything. That matters for:
- No-browse retrieval pipelines — RAG systems that ingest text but cannot make follow-up HTTP requests
- Embedded agents — running in environments without arbitrary network access
- One-shot context — when an LLM needs the whole knowledge base in its context window up front
- Training and fine-tuning ingestion — when content goes into a model build pipeline rather than a runtime fetch
For browsing agents, llms-full.txt is harmless overhead (they can choose not to fetch it). For non-browsing agents, it is the difference between "the agent has your docs" and "the agent has nothing about you."
Format
Plain text or markdown. Standard convention is to start with the same header as llms.txt, then concatenate every page in your llms.txt directory, separated by clear headings or --- separators.
Example skeleton:
# Your Service Name
> Brief one-line description.
## Overview
Your Service does X, Y, Z.
---
### /docs/getting-started
(full markdown content of the getting-started page)
---
### /docs/api-reference
(full markdown content of the API reference)
(...continues for every page)
How agents discover it
Two ways:
1. <link rel="alternate"> in your HTML head
<link rel="alternate" type="text/plain" href="/llms-full.txt" title="Full content for LLMs">
2. Reference it from /llms.txt
Most llms.txt files end with an "Optional" section. Add the companion there so agents reading the directory know the full-content version exists.
## Optional
- [Full content](/llms-full.txt) — entire site text in one file
How to generate it
Most modern docs platforms (Mintlify, Docusaurus with plugins, Hugo with templates) can auto-generate llms-full.txt from your source. If you maintain docs manually, a simple build step that concatenates your markdown source files works fine.
Regenerate when content changes. A stale llms-full.txt is worse than none — it gives agents outdated information about your service. Wire generation into the same CI step that builds your docs.
Tradeoffs
- Pro — A single fetch gives consumers everything; no crawling needed
- Pro — Strong AEO signal: ChatGPT, Perplexity, and Claude all weight machine-readable corpus availability
- Pro — Works for agents without browsing capability (RAG pipelines, embedded agents)
- Con — Larger response: tens of KB to several MB depending on docs size
- Con — Must be regenerated when content changes
- Con — Concatenates everything, including content you might prefer agents fetch piecemeal (paid docs, gated content)
What agentgrade checks
- llms-full.txt found —
/llms-full.txtresponds 200 with at least 20 characters of non-HTML content that does not look like a soft-404 - llms-full.txt linked from HTML (optional) —
<link rel="alternate" type="text/plain" href="/llms-full.txt">is present in your homepage<head>
The check is scored at weight 1 — lower than llms.txt itself at weight 2, reflecting that llms-full is recommended companion content rather than a primary directory file.
Real-world examples
- docs.anthropic.com/llms-full.txt — Anthropic's full Claude docs in one file
- agentgrade.com/llms-full.txt — our own dogfood version, generated from this knowledge base
- The original llmstxt.org spec describes both files as the canonical pair
Spec maturity
Emerging standard, growing fast. Defined alongside llms.txt at llmstxt.org. Adoption is rising, driven by AI-first docs platforms (Mintlify auto-generates it) and the spread of no-browse retrieval workflows. Sites that ship both files now position themselves as authoritative sources for AI answer engines.
Learn more
- llmstxt.org — Specification
- llms.txt — The companion directory file