CLI for modelschemas — live per-endpoint request/response JSON Schemas and model metadata for AI providers (OpenAI, Anthropic, Gemini, xAI Grok, ElevenLabs, OpenRouter, FAL).

The published bin is built JavaScript and runs on Node. No TypeScript loader.

bunx modelschemas models list --activity chat
bunx modelschemas schema get anthropic v1/messages --kind input
bunx modelschemas validate anthropic/v1/messages payload.json

login [--name <n>] [--delegated] [--api-key]   register this machine
whoami                                         show identity, grants, usage
models list|get                                model catalog
schema get <provider> <endpointId>             one self-contained JSON Schema
validate <provider/endpointId> <file>          validate a payload (exit 2 = invalid)
estimate <provider> <model> [body.json]        evaluate the stored rate card
                                               ({request?,usage?} or usage object)
changes [--since <epoch>]                      changelog feed
subscribe <url>                                webhook subscription
pull <selection...>                            generate schema/type modules
update                                         refresh pulled modules
verify                                         re-hash pulls against pinned content
                                               addresses + provenance

--json everywhere (default when piped), so the CLI is agent-friendly. Output of pull is committed; verify proves the service still serves bytes matching each pinned content hash.

Docs: modelschemas.com/docs · Source: github.com/modelschemas/modelschemas