Audited October 16, 2026 · scoring 100-point-v4
MotherDuck
C-MotherDuck’s public docs score 70.0% (C-): content delivery 50/50, discovery 16/25, agent capabilities 3/25. 5 of 10 checks earn full points. Largest gap: MCP server.
70.0% · 70/100 · AI Agent Readiness Score
Content delivery · 50/50
Discovery · 16/25
Agent capabilities · 3/25
Extra credit · 1/2 · cannot raise the score above 100
docs platform: Docusaurus 3.10.2 (unscored) · scoring 100-point-v4
What to fix first
5 checks lose points. They are ranked by the points a fix would add back.
- 1 +15 MCP server
Offer a Model Context Protocol server that answers without sign-in, so agents can search the docs through tools.
- 2
Publish an OpenAPI description at a discoverable address when the product has an API.
- 3
Add a canonical link, a meta description, Open Graph tags, and JSON-LD to the server HTML.
- 4 +4 llms.txt
Publish an llms.txt index under 50,000 characters whose links point to Markdown pages.
- 5
Add agent discovery signals such as Content Signals, an agent skills index, or an MCP server card.
Every check, with evidence
Each check shows what the audit found and the points it earned. Evidence links open the exact URL the audit requested.
Content delivery · 50/50
Content in server HTML
15/15 PASSAgents and crawlers that fetch a page without a browser read only the HTML response. This check looks for the page’s main content in that response.
Found: The server HTML has 2,203 characters of main content; 250 are required.
Markdown mirror
15/15 PASSMarkdown is the cheapest format for an agent to read. This check looks for a Markdown version of this page and confirms that it is the same page.
Found: method: accept-negotiation (Vary: Accept); 4,445 characters
- Markdown version of this page (10/10)
- Markdown matches the page (5/5)
Evidence: https://motherduck.com/docs/getting-started/ (opens in a new tab) · HTTP 200
Page fits one agent fetch
10/10 PASSAgent tools commonly cut off long responses. This check measures the text an agent receives from this page.
Found: The Markdown version has 4,445 characters.
AI crawler access
10/10 PASSAI search and user-triggered fetch agents read docs when a person asks a question. This check reads robots.txt rules for those agents.
Found: 0 of 11 AI search and user-triggered fetch agents are blocked in robots.txt.
Discovery · 16/25
llms.txt
11/15 PARTIALllms.txt is an index that points agents to the documentation. This check validates it, measures it, and samples its links.
Found: 14 sections, 54 links; 17,023 characters; no .md links to sample
- Valid llms.txt index (7/7)
- llms.txt under 50,000 characters (4/4)
- llms.txt links return Markdown (0/4, missing)
Evidence: https://motherduck.com/llms.txt (opens in a new tab) · HTTP 200
Sitemap and robots.txt
5/5 PASSA sitemap lists every page, so crawlers do not depend on navigation to find them.
- robots.txt present (1/1)
- Sitemap declared (2/2)
- Sitemap valid XML (2/2)
Page metadata
0/5 FAILStructured page metadata in the server HTML helps machines identify the page and its preferred address.
- JSON-LD in raw HTML (0/1.25, missing)
- Canonical link (0/1.25, missing)
- Meta description (0/1.25, missing)
- Open Graph (0/1.25, missing)
Agent capabilities · 3/25
MCP server
0/15 FAILA Model Context Protocol server lets agents search or use the docs through tools.
Found: server card declared, not verified live
Evidence: https://motherduck.com/.well-known/mcp.json (opens in a new tab) · HTTP 200
API description
0/5 FAILA machine-readable API description lets coding agents build API calls without guessing from prose.
- OpenAPI document (0/3, missing)
- API catalog (.well-known) (0/2, missing)
Agent interoperability signals
3/5 PARTIALEmerging standards that help agents discover and use a site.
- Content Signals (1/1)
- Agent skills index (1/1)
- OAuth discovery (0/1, missing)
- Link headers (0/1, missing)
- Web Bot Auth directory (0/1, missing)
- MCP server card (1/1)
- Agent login instructions (auth.md) (0/1, missing)
Extra credit · 1/2
Extra credit can restore lost points. It cannot raise the score above 100.
llms-full.txt
1/1 PASSllms-full.txt combines the docs into one file for tools that load a whole documentation set.
Found: 31.4 KB, ~8,031 tokens (approx)
Evidence: https://motherduck.com/llms-full.txt (opens in a new tab) · HTTP 200
Docs AI assistant
0/1 FAILAn assistant on the docs site answers readers’ questions. It helps people, not external agents.
Found: none detected
Archived reading test
MotherDuck had a reading test on 2026-08-19, under the method the series used until October 16, 2026. Three AI models read the docs and attempted five first-hour developer jobs. It scored 96.2% (A). That score is not comparable with this one.
Show the score
Paste this into a readme:
[](https://docsforagents.com/reports/motherduck-docs-ai-agent-readiness/)Method note
This audit requested the public MotherDuck docs on October 16, 2026 and scored what the server returned. No AI model read the docs or scored anything. The audit did not run JavaScript, sign in, or create an account. Scoring version 100-point-v4.
Related reports
- Hermes Agent Grade F · 57.0%
- Overture Maps Grade F · 49.0%
- Airbyte Grade C+ · Docusaurus 3.10.1
Audit another docs site.
Nominate a docs site