As AI agents become autonomous network participants, the absence of standardized, verifiable identity creates systemic vulnerabilities including impersonation, unauthorized resource access, and regulatory non-compliance. This paper presents the Elpis Protocol, an infrastructure-level cryptographic identity framework anchored on the XRP Ledger (XRPL). Elpis operates transparently within the agent's execution environment, requiring zero application-level changes. The protocol introduces a Three-Tier Identity model (Root CA, Provider CA, Agent Certificate), a Four-Level Escalation mechanism for graduated incident response, and a Two-Tier Reputation system combining on-chain attestations with off-chain behavioral scoring. A reference implementation demonstrates sub-5ms validation overhead via a transparent signing proxy that attaches cryptographic identity headers to all outbound requests. Elpis is fully compatible with existing OAuth 2.0, mTLS, and OpenID Connect infrastructure.