SCL for AI Agent Identity
Definition
SCL is the deterministic, cryptographic identity protocol for AI run declarations.
SCL defines how deterministic identity for AI is assigned to agent declarations. The protocol produces a SHA-256 identity from the canonical JSON representation of AI run declarations.
Identity for AI Agents
AI agents generate declarations that require stable, implementation-independent identity. SCL assigns this identity exclusively through a deterministic process: the declaration is transformed into canonical JSON, and the SHA-256 identity is derived exclusively from canonical JSON bytes.
Canonical JSON is the representation. Canonical JSON bytes are the sole input to SHA-256.
The resulting SHA-256 identity serves as the fixed reference for the AI run declaration. This identity remains identical across all conforming implementations when the input declaration bytes are identical.
This requirement exists independently of SCL and is recognized across the industry. Industry discussions increasingly recognize the need for robust identity mechanisms for AI agents and software. For example, recent work from NIST highlights the importance of advancing adoption of software and artificial intelligence agent identity and authorization.
Accelerating the adoption of software and artificial intelligence agent identity and authorization remains a priority area.
— NIST Concept Paper, February 2026
SCL addresses this requirement by providing a concrete, deterministic identity layer based solely on canonical JSON bytes.
Core Properties
- Deterministic identity: identical declarations always produce the same SHA-256 identity.
- Implementation independence: the identity depends only on the canonical JSON representation.
- No execution semantics: SCL assigns identity only and defines no behavior or outcomes.
- No normalization or interpretation: identity is derived from exact canonical JSON bytes only.
Transformation Pipeline
The deterministic process follows the SCL:V1 specification:
bytes → AST → canonical JSON → SHA-256
This pipeline ensures that every AI agent declaration receives a unique, reproducible SHA-256 identity. The SHA-256 identity is derived exclusively from canonical JSON bytes.
Use Cases
- Referencing specific AI agent declarations in logs, audits, or registries (see the Golden Documents for concrete examples).
- Binding authorizations or attestations to a fixed AI run declaration identity.
- Enabling consistent tracking of agent declarations without reliance on runtime systems.
SCL provides the deterministic identity layer required for AI agent systems that demand verifiable, stable references. These identities can be confirmed with the reference engine.
See also: SCL in multi-agent systems.