Introduction

What is IDProva?

IDProva is an open protocol for establishing verifiable identity, scoped delegation, and auditable action tracking for autonomous AI agents. Built on the W3C Decentralized Identifier (DID) standard, IDProva introduces:

• did:idprova: — A DID method designed specifically for AI agents
• Delegation Attestation Tokens (DATs) — Scoped, time-bounded, revocable permission tokens
• Action Receipts — Hash-chained, signed audit records that map to compliance frameworks

Why do agents need their own identity layer?

AI agents are proliferating — but they lack any standard way to prove who they are, what they’re authorised to do, or on whose behalf they act.

Current Approach	Limitation
API keys / bearer tokens	No identity semantics; shared secrets; no delegation hierarchy
OAuth 2.0 client credentials	Designed for applications, not autonomous agents; no agent metadata
SPIFFE/SPIRE	Workload identity only; no delegation model; no agent-specific semantics
Custom per-vendor solutions	Vendor lock-in; no interoperability; no audit standard

Without a universal identity layer, the agent ecosystem faces systemic risks:

• Agents impersonating other agents
• Uncontrolled privilege escalation through opaque delegation chains
• Regulatory non-compliance due to absent audit trails
• Inability to establish trust across organisational boundaries

Design Principles

IDProva is guided by these principles, in priority order:

1. Security First — All identity claims and delegations are cryptographically verifiable
2. Post-Quantum from Day One — Hybrid Ed25519 + ML-DSA-65 (FIPS 204) signatures
3. Progressive Trust — Agents start at L0 and earn trust through verifiable mechanisms
4. Minimal Disclosure — Agents disclose only what’s necessary for each interaction
5. Decentralised by Default — Self-hosted registries are first-class citizens
6. Protocol Composability — Layers on MCP, A2A, HTTP — no new transport required
7. Regulatory Alignment — Audit trails satisfy ISM, SOC 2, and NIST 800-53 out of the box
8. Developer Experience — Basic functionality implementable in a weekend

Three Pillars

Identity

Every agent gets a DID Document with public keys, metadata (model, runtime, configuration hash), trust level, and capabilities. The DID Document is the root of trust.

Delegation

Authority flows through Delegation Attestation Tokens — signed, scoped, time-bounded, and chainable. Verifiers trace any delegation back to its root principal (human or org).

Audit

Every action produces a signed Action Receipt, hash-chained into a tamper-evident log. Each receipt references the DAT that authorised it — complete attribution from action to principal.

Relationship to Existing Standards

IDProva builds on established standards:

• W3C DIDs v1.0 — did:idprova: is a conforming DID method
• JWS (RFC 7515) — DATs use JWS Compact Serialization
• JWT (RFC 7519) — DAT payloads follow JWT claim conventions
• FIPS 204 (ML-DSA) — Post-quantum signature component
• SPIFFE — Inspired the DID URI format
• MCP / A2A — Protocol binding specifications provided

SDKs & Tools

IDProva ships with SDKs for three languages and a CLI, all powered by the same Rust core:

SDK	Install	Bindings
Python	pip install idprova	PyO3 (native Rust)
TypeScript	npm install @idprova/core	napi-rs (native Rust)
Rust	idprova-core = "0.1"	Reference implementation
CLI	cargo install idprova-cli	Direct binary

All SDKs expose the same API surface: KeyPair, AID, DAT, Scope, TrustLevel, and ReceiptLog.

Next Steps

• Quick Start — Create your first agent identity in under 5 minutes
• Python SDK — Python SDK guide
• TypeScript SDK — TypeScript SDK guide
• Concepts: Identity — Deep dive into Agent Identity Documents
• Protocol Specification — Full technical reference