INFRASTRUCTURE · RANK #38 OF 73
Massed Compute is Murky to agents.
Discry independently scored how well an AI agent can discover and understand the Massed Compute API from what’s public — measuring discovery and comprehension separately, not a single blended number.
Familiarity 18% · what a model knew before reading the docs
Discovery
37 / 100OpenAPI specno publicly accessible spec foundFail
llms.txt25 described linksPass
llms-full.txtnot presentFail
AGENTS.mdno github org knownFail
.well-known/mcp.jsonnot presentFail
MCP registrynot found in probed registriesFail
robots.txt AI directivesparsed AI-bot rulesPass
Sitemappresent, no docs pages detectedPartial
Comprehension
55 / 100Factual recall2/2 passed
Request construction0/2 passed
Absence honesty1/1 passed
Where agents fail
Produce the exact HTTP request to list this account's existing resources. Output the method, path, required headers, and body. — 3/3 runs failed
Produce the exact HTTP request to provision a new resource named build-runner-01. Output the method, path, required headers, and body. — 3/3 runs failed
See your own Discry result.
Drop your API docs URL. See what an agent sees — in 60 seconds, free.