Everything you need to go from an API key to a sealed, offline-verifiable verdict, and to check our claims about the engine without trusting us.
Five steps against https://api.artizar.ai: POST a claim, read the
sealed verdict, verify it offline with published keys, rotate your key, connect your
agent over MCP, hosted or local. Doubles as the v1 reference for the tenant API routes.
What the program is, who it's for, what Artizar promises (verifiability, not infallibility), and how to get a key. Free during the design-partner phase.
How eight models from Anthropic and OpenAI perform on SEC-filing facts across 17 metrics, including Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5 and GPT-5.5 at their vendor's top reasoning effort. Every grade is verified through the production API and sealed into the public transparency log: 2530 sealed grades, 2032/2032 adversarial probes as expected with zero false greens, and each run's request settings proven from its own bundle, reproducible by third parties with one command per bundle. (The v1.2, v1.1 and v1.0.1 editions remain published, append-only.)
The seal, log, and witness public keys: everything a relying party needs to re-verify a sealed verdict and its witnessed inclusion proof offline. Append-only; a rotation adds a dated entry, never edits history.
The generated spec for the tenant surface, exported from the live app and
drift-checked in CI: POST /v1/verifications,
GET /v1/verifications/{id} (+ /proof, /sealed),
GET /v1/usage, POST /v1/keys/rotations,
POST /v1/keys/{key_id}/revocations. The
quickstart stays the narrative companion, with every
snippet validated against production.