Roboflow
Last scanned docs.roboflow.comGitBook15 pages tested
AI agents can use most of these docs. A few fixes would make them reliable. Roboflow scores 89 out of 100 (grade B), passing 18 of 27 checks. 9 still need work.
Developer docs
Kind of site
18/27
Checks passed
4
Partly passed
5
Failed
Access
Content discoverability4 passed1 failed2 partly
llms.txt fits in an agent's context
llms.txt is 144,852 characters (exceeds 100,000 threshold; will be truncated by most agents)
Your llms.txt is 144,852 characters and will be truncated by all major agent platforms. Split into a root index linking to section-level llms.txt files, each under 50,000 characters.
Check your fix:
curl -s https://docs.roboflow.com/llms.txt | wc -cprints less than 50000.llms.txt follows the standard structure
llms.txt contains parseable links but doesn't fully follow the proposed structure: https://docs.roboflow.com/llms.txt: No blockquote summary found
Your llms.txt contains parseable links but doesn't follow the standard structure. Add an H1 title as the first line and a blockquote summary (lines starting with >) to improve agent parsing.
Check your fix: llms.txt starts with a
# Title, then a> summaryline, then##sections of links.Pages point agents to llms.txt
llms.txt directive found in HTML of 15 of 15 sampled pages, but buried deep in the page (past 50%)
An llms.txt directive was found in the HTML of some pages but is missing from others, or is buried deep in the page. Ensure the directive appears near the top of every documentation page.
Check your fix:
curl -s https://docs.roboflow.com/agents/agents | grep -n llms.txtfinds the link near the top of the HTML.llms.txt exists
llms.txt found at https://docs.roboflow.com/llms.txt
llms.txt links work
All 15 same-origin sampled links resolve (808 total links)
llms.txt links point to Markdown
15/15 same-origin sampled links point to markdown content (100%)
Markdown pages point agents to llms.txt
llms.txt directive found in markdown of all 15 sampled pages, near the top of content
Markdown availability2 passed
Pages are available as Markdown
15/15 sampled pages support .md URLs (100%)
Server returns Markdown when asked
15/15 sampled pages support content negotiation (100%)
Page size and truncation1 passed1 failed2 partly
Content starts near the top
8 of 15 sampled pages have content starting past 50% (worst 76%)
8 of 15 pages have content starting past 50% of the converted output. Agents may never see the documentation content. Reduce navigation, breadcrumb, and sidebar markup that precedes the content area.
Markdown pages fit in context
1 of 15 pages between 50K–100K chars (max 50K)
1 of 15 markdown pages are between 50K and 100K characters. These may be truncated on some agent platforms or routed through summarization. Consider splitting large pages.
HTML pages fit in context
1 of 15 sampled pages convert to 50K–100K chars (max 1572K HTML → 60K markdown (98% boilerplate))
1 of 15 pages convert to 50K-100K characters of markdown. Review pages for reducible boilerplate (navigation, serialized tabbed content). Consider providing markdown versions as a smaller alternative path for agents.
Content is in the HTML without JavaScript
All 15 sampled pages contain server-rendered content
Content structure2 passed1 failed
Headings inside tabs name their variant
1 of 1 page(s) with tab headers don't distinguish between variants (e.g. "kotlin" repeats across 2 tab groups)
Over 50% of headers are generic across tab variants. When serialized, agents cannot tell which section belongs to which variant.
Tabs don't bloat the page
5 tab group(s) across 1 of 15 sampled pages; all serialize under 50K chars
Code blocks are well-formed
All 66 code fences properly closed across 16 pages
URL stability2 passed
Missing pages return a real 404
All 15 sampled pages return proper error codes for bad URLs
Redirects stay on the same host
No redirects detected across 15 sampled pages
Content health2 passed1 failed
Markdown matches the HTML page
2 of 15 pages have substantive content differences between markdown and HTML (avg 8% missing)
2 pages have substantive content differences between markdown and HTML (avg 8% missing). If unintentional, agents are getting outdated content; regenerate markdown from source or fix the build pipeline.
llms.txt covers the whole site
llms.txt covers 99% of 807 sitemap doc pages; 8 llms.txt links not in sitemap (may indicate stale links or incomplete sitemap)
Cache headers allow timely updates
All 16 endpoints have appropriate cache headers
Authentication and access1 passed
Gated docs offer another way in
All docs pages are publicly accessible; no alternative access paths needed
Docs are readable without logging in
All 15 sampled pages are publicly accessible
Crawler access1 passed1 failed
Sitemap lists the docs pages
The sitemap at https://docs.roboflow.com/sitemap.xml lists 161 of the 816 docs pages found through llms.txt and page links.
Generate the sitemap from the same source as the docs, so every page is listed. Agents and crawlers use it to find pages that navigation does not link.
Check your fix:
curl -s https://docs.roboflow.com/robots.txt | grep -i sitemapnames a sitemap, and that sitemap lists the pages the report names as missing.Crawlers and AI assistants may read the docs
robots.txt lets Googlebot, Bingbot, OAI-SearchBot, PerplexityBot, Claude-SearchBot and Claude-User read the docs.
Freshness
Maintenance3 passed
Pages say when they were last changed
14 of 15 sampled pages state when they were last changed.
Links and section anchors work
All 40 links and 15 section anchors checked on the sampled pages work. 2 more could not be verified (timeouts or blocked requests) and are not counted.
Sitemap entries still exist
All 10 sitemap entries sampled still exist.
Consistency with the API specNot applicable
Endpoints in the docs exist in the API spec
No OpenAPI description was found, so the docs could not be compared with a spec.
CurrencyNot applicable
Deprecated operations are marked as such
No OpenAPI description was found, so deprecation notices could not be compared with a spec.
Additional checks
Not part of the score: optional extras that help agents use your docs.
MCP server is discoverable
An MCP server answers at https://mcp.roboflow.com/mcp; agents need to sign in to use it.
llms-full.txt is available
llms-full.txt found at https://docs.roboflow.com/llms-full.txt (638,264 characters).
Agent skills are published
No agent skills index was found at /.well-known/agent-skills/index.json.
Publish an Agent Skills index at /.well-known/agent-skills/index.json with a SKILL.md for your main tasks, so agents can load task instructions without reading every page.
About this scan
- Scanned
- Oct 21, 2026
- Pages tested
- 15 of 808 found
- Platform
- GitBook
- Kind of site
- Developer docs
- AFDocs score, on its own
- 89 (B), 14 of 23 checks pass
- Based on 15 sampled pages, chosen deterministically from the sitemap and llms.txt. Pages outside the sample were not read.
- Pages were fetched without running JavaScript, the way most AI agents read the web.
- Answerability, where an AI agent answers real questions written from these pages, runs in the open-source scanner with your own API key. This web score is made of Access and Freshness.
Methodology 2026-09.6 · AFDocs 0.20.0
Pages we sampled (15)
- https://docs.roboflow.com/agents/agents
- https://docs.roboflow.com/datasets/manage/manage-images
- https://docs.roboflow.com/deployment/monitoring-and-analytics/vision-events/send-events
- https://docs.roboflow.com/deployment/self-hosted/enterprise/kubernetes
- https://docs.roboflow.com/models/evaluate/view-trained-models
- https://docs.roboflow.com/models/train/roboflow-universe-checkpoints
- https://docs.roboflow.com/reference/changelog/changelog/april-2026
- https://docs.roboflow.com/reference/changelog/changelog/january-2026/gemini-3-flash-in-workflows
- https://docs.roboflow.com/reference/changelog/changelog/march-2025/deprecating-python-sdk-inference
- https://docs.roboflow.com/reference/changelog/changelog/october-2026/rf-detr-1.11-export-to-openvino-litert-and-core-ai
- https://docs.roboflow.com/workflows/blocks/blocks-governance
- https://docs.roboflow.com/workflows/blocks/blocks/data-storage/roboflow-dataset-upload
- https://docs.roboflow.com/workflows/blocks/blocks/run-a-model/qwen2-5-vl
- https://docs.roboflow.com/workflows/blocks/blocks/visualize-predictions
- https://docs.roboflow.com/workflows/developer-guide/developer-guide/kinds/detection
Show your score
Add the badge to your docs or README. It updates each time the site is scanned.
Markdown
[](https://documentation.ai/agent-score/roboflow)HTML
<a href="https://documentation.ai/agent-score/roboflow"><img src="https://documentation.ai/agent-score/badge/roboflow.svg" alt="Agent Readiness Score: 89 (B)" height="20"></a>Agent-ready docs by default
Live in under 5 min · No credit card required