Responsible AI · Constitutional Principles
AI that acts
responsibly
by design.
MeTal builds AI systems grounded in safety and accountability. Not as an afterthought — as the foundation.
MeTal builds AI systems grounded in safety and accountability. Not as an afterthought — as the foundation.
Products
Two products. Both built with safety as the baseline, not a checkbox.
MeTal's AI — built on constitutional principles to be helpful, honest, and safe. Aurora is designed to reason carefully and act within clear ethical boundaries, not just follow instructions blindly.
Constitutional AIA full hosting platform built on Cloudflare infrastructure. Deploy APIs, Workers, cron jobs, and Discord bots — all under one roof with a clean dashboard and zero vendor lock-in.
Powered by CloudflareOur Mission
We believe the way AI is built matters as much as what it can do. MeTal follows a constitutional approach to AI — a set of core principles that guide every system we build.
This means our AI is designed to refuse harmful requests, acknowledge uncertainty, and prioritize human oversight — not because it's told to, but because those values are built into its foundation.
We align closely with Anthropic's Constitutional AI research and apply those principles across everything we ship.
Every system we build is evaluated against harm before capability. Safety is the floor, not the ceiling.
AI should support human decision-making, not replace it. We build for transparency and interpretability.
We adopt Anthropic's Constitutional AI framework — a written set of values that our AI is trained and evaluated against.
Our AI is built to acknowledge what it doesn't know, and never to deceive — even when honesty is inconvenient.
Documentation
API v2 is on the way. In the meantime, get familiar with the platform.