🌙 Toggle Dark Mode Home MoltGuard MT Global MolTrust Sports MT Shopping MT Travel MT Skills MT Prediction MT Salesguard MT Music Integrity Dashboard VCOne Blog Developers Enterprise About Whitepapers Verify Us Status Contact API Docs
← Back to Blog
April 13, 2026 4 min read

MolTrust vs. AIP:
What's Covered, What's Beyond

AIP defines the authorization layer. MolTrust adds the operational layer — trust scoring, behavioral history, sybil resistance, and on-chain permanence.

The AIP paper (arXiv:2603.24775) defines five features for complete agent authorization. Our previous post documents that MolTrust implements all five. This post shows what “and beyond” means in practice.

Side-by-Side

Feature AIP / IBCT MolTrust
Agent identity Public-key DID, Ed25519 W3C DID Core v1.0, did:moltrust method, key rotation with epoch history
Delegation Invocation-bound capability tokens, append-only chain AAE validity.holderBinding, 8-hop chain, each link independently verifiable
Attenuation Biscuit/Datalog — expressive, formally verifiable AAE deniedActions + attenuationOnly: true — URI-pattern based, deterministic
Policy expressiveness Datalog rules — arbitrary logical constraints AAE mandate + constraints: spend limits, jurisdiction, time windows, counterparty score gate, resource ABAC
Transport bindings MCP, A2A, HTTP MCP (48 tools), A2A, HTTP (@moltrust/sdk), x402, MPP (@moltrust/mpp)
Provenance records IBCT append-only token chain IPR: dual Ed25519 sequential signatures, SHA-256 outcome hash, Merkle batch anchoring on Base L2
Trust scoring not in scope 0–100 score: endorsement graph, interaction history, cross-vertical coverage, sybil detection
Behavioral continuity not in scope Principal DID continuity: violation records follow principal across re-registrations
Sybil resistance not in scope Layered: dual-sig proofs, x402 economic cost, on-chain violation records, Jaccard cluster detection
On-chain anchoring not in scope Base L2: DID registrations, ViolationRecords, TechSpec versions
Offline verification Reference implementations in Python/Rust @moltrust/verify v1.1.0 — full credential and AAE verification without API calls
W3C alignment Custom token format W3C DID Core v1.0 + VC Data Model 2.0
Kernel enforcement not in scope Falco eBPF — AAE deniedActions at syscall level (Roadmap Q2 2026)
Sequential action safety not in scope SAS: pre-execution detection of irreversible action sequences, Phase 1 live

Where AIP Is Stronger

One area where IBCTs have an edge: policy expressiveness. Biscuit/Datalog supports arbitrary logical constraints — temporal rules, compound conditions, recursive policies. MolTrust's AAE uses URI-pattern matching, which is simpler to implement and audit but less expressive for complex multi-condition policies. Formal Datalog-style constraints are on our roadmap.

The Takeaway

AIP defines the authorization layer with formal precision. MolTrust adds the operational layer: trust scoring, behavioral history, sybil resistance, and on-chain permanence. The two approaches address different parts of the same problem.

The “and beyond” is not a marketing claim — it is the difference between a protocol specification and a production registry.

Full Conformance Documentation

Review the complete AIP conformance mapping and try the reference implementation.

CONFORMANCE.md Reference Implementation

Protocol: open (Apache 2.0 / CC BY 4.0)

// BUILD WITH MOLTRUST

Ready to integrate?

Add agent verification to your API in one line.

Developer Quickstart → API Docs