Browserbase scored 94/100 on Mintlify's public score report across 29 published checks. This Documentation benchmark shows where agents can use these docs cleanly and where the biggest points of friction still are.
This report passes 25 of 29 checks, with 1 warnings, 1 failed checks, and 2 skipped checks still shaping the next score lift.
Documentation quality now affects implementation speed, support load, and whether an agent or buyer can stay self-serve. This benchmark shows where the docs already work well and where they still introduce friction.
Good documentation is discoverable, readable in plain text, clear about structure and instructions, and specific enough in examples and next steps that real workflows can keep moving.
These are the clearest signals behind the current score. Start with the strengths that already support discovery and implementation, then use the gaps below to prioritize the fixes most likely to improve AI readability and self-serve onboarding.
For the broader pattern behind these wins, review AI-readable documentation and compare this page with the full docs benchmark leaderboard.
If you want to improve the system behind these gaps, start with API docs automation, the OpenAPI to CLI generator, and this guide to making docs AI-readable.
Open each category to review the checks behind the score and see exactly where agents still hit friction.
Can an agent find the right starting point?
llms.txt found at 1 location(s)
llms.txt follows the proposed structure (H1, blockquote, heading-delimited link sections)
llms.txt is 25,251 characters (under 50,000 threshold)
All 15 same-origin sampled links resolve (167 total links)
15/15 same-origin sampled links point to markdown content (100%) (4 external links excluded)
llms.txt directive found in all 15 sampled pages, near the top of content
Can an agent fetch a clean text version of the docs?
15/15 sampled pages support .md URLs (100%)
15/15 sampled pages support content negotiation (100%)
Can an agent read the page without losing important context?
All 15 sampled pages contain server-rendered content
All 15 pages under 50K chars (median 5K, max 13K)
15 of 15 sampled pages convert to over 100K chars (max 563K, 32% boilerplate) 15 of 15 pages convert to over 100K characters of markdown. Reduce inline CSS/JS, break large pages, or provide markdown versions as a smaller alternative.
Content starts within first 10% on all 15 sampled pages (median 4%)
Can an agent understand how the page is organized?
25 tab group(s) across 10 of 15 sampled pages; all serialize under 50K chars
10 page(s) with tabs found, but no section headers inside tab panels to evaluate
All 25 code fences properly closed across 16 pages
Can an agent trust that URLs will stay predictable?
All 15 sampled pages return proper error codes for bad URLs
No redirects detected across 15 sampled pages
Can an agent tell whether the docs are fresh and trustworthy?
llms.txt covers 100% of 161 sitemap doc pages; 2 llms.txt links not in sitemap (may indicate stale links or incomplete sitemap)
4 of 15 pages have minor content differences between markdown and HTML 4 pages have minor content differences between their markdown and HTML versions. Review for formatting variations.
All 16 endpoints have appropriate cache headers
Can an agent access the docs without getting blocked?
All 15 sampled pages are publicly accessible
All docs pages are publicly accessible; no alternative access paths needed
Can a long-context agent ingest a full docs snapshot directly?
Found llms-full.txt.
llms-full.txt size is within the expected range.
llms-full.txt has a recognizable markdown structure.
llms-full.txt links resolve successfully.
Do these docs publish explicit operating guidance for agents?
Found an agent skill definition.
Can an agent discover first-class tools instead of scraping pages?
Found an MCP server.
The MCP server exposes tools.
Save the report to a work email now, or tell us what kind of ongoing monitoring would be useful for this docs site.
Compare related benchmark reports, then move into the guides and tools that improve documentation benchmark scores.
98/100 score in DevTools. Open the full documentation benchmark report.
96/100 score in DevTools. Open the full documentation benchmark report.
95/100 score in DevTools. Open the full documentation benchmark report.
Compare this documentation benchmark report with other public docs benchmark results.
Score your own API docs or developer docs and open the full public report.
See how docs, SDKs, CLIs, and agent-facing outputs can ship from one OpenAPI-driven system.
Use the CLI page when the benchmark gaps are really onboarding and implementation gaps.
Use this guide when you want to improve the same categories this benchmark scores.
DocsAlot helps teams improve help centers, developer docs, API docs, and CLI docs so they are easier for humans to use and easier for agents to read.