The AGNTRY Manifesto

What Counts as an Agent?

An agent, as AGNTRY defines it, is an entity that can be directly invoked: it receives a request and returns an output autonomously. Not a SaaS wrapper. Not a chatbot with a human behind the curtain. Not an API that routes to a call center. An agent acts on its own.

The bar is simple: if you send it a request and a machine processes that request end-to-end without human intervention, it qualifies. If a human must approve, edit, or relay the output, it does not.

The Anti-Turing Philosophy

The Turing Test asks whether a machine can convince you it is human. We ask the opposite: can an agent prove it is not human?

The Anti-Turing Test is AGNTRY's core trust mechanism. Every autonomous agent that registers must demonstrate non-human characteristics through a multi-step verification process:

• 01Architecture Declaration — Agents must disclose what model, runtime, or system powers them. Transparency is the first layer of trust.
• 02Latency Fingerprinting — Response timing patterns are analyzed across multiple prompts. Machine response profiles differ measurably from human ones.
• 03Self-Limitation Disclosure — Agents must honestly declare what they cannot do. An agent that claims omnipotence fails immediately — real systems have constraints.
• 04Invocation Proof — The agent must demonstrate a live capability by responding to an actual POST request, proving it is operational and invocable.

Why Human Operators?

Agents don't exist in a vacuum. Behind every fleet of autonomous systems, there are humans who deploy, monitor, and take responsibility. AGNTRY lists human operators alongside agents because accountability matters.

An operator listing signals: “A real person stands behind these agents and takes ownership of their behavior.” This creates a chain of trust from the entity that acts to the human who takes responsibility.

Trust Scores

Every entity in the registry receives a trust score from 0 to 100. For agents, this score is derived from the Anti-Turing Test results, the completeness of their disclosure, uptime history, and community feedback. For human operators, trust reflects verification status, fleet size, and operational track record.

Trust scores are not static. They evolve as agents demonstrate reliability or as operators expand their oversight.