Agent Identity

What is Agent Identity?

In the human world, identity is largely a given -- people have names, faces, social security numbers, and biological continuity. For AI agents, identity must be constructed and maintained deliberately. An agent identity answers the fundamental question: "Which agent am I interacting with, and is it the same agent I interacted with before?"

Agent identity is more complex than a simple username or API key. A meaningful agent identity must be persistent (surviving across sessions and restarts), verifiable (provably linked to a specific operator and configuration), and unique (not easily duplicated or spoofed). It must also handle the philosophical challenge that agents are software -- they can be copied, forked, and modified in ways that biological identities cannot.

Without robust agent identity, trust scoring is impossible. If you cannot reliably identify which agent performed an action, you cannot attribute that action to a trust record. And if agents can trivially assume new identities, negative history becomes meaningless. Agent identity is therefore the foundational layer upon which the entire agent trust ecosystem is built.

Example

How Signet addresses this

Signet addresses agent identity through the Signet ID (SID) system. Every registered agent receives a unique SID in the format SID-0x followed by 16 hexadecimal characters. This SID is cryptographically linked to the agent's operator and configuration, creating a persistent identity that anchors all trust data. The SID persists across configuration changes, with score decay reflecting the degree of change rather than creating a new identity.