We have published a research paper describing the Attestix architecture and its approach to solving the AI agent trust problem.
Abstract
As AI agents proliferate across enterprise workflows, the need for verifiable identity, auditable actions, and regulatory compliance becomes critical. Attestix presents a comprehensive attestation infrastructure that provides cryptographic identity, W3C Verifiable Credentials, capability-based delegation, and EU AI Act compliance tooling through a 47-tool MCP server.
Key Contributions
-
Universal Agent Identity Token (UAIT) - A bridge format connecting MCP server names, A2A agent cards, and W3C DIDs into a single identity namespace
-
Ed25519Signature2020 Credentials - Full W3C Verifiable Credentials implementation with JSON-LD context and revocation support
-
UCAN Delegation Chains - Capability-based delegation using signed JWT tokens with scoped permissions and time bounds
-
EU AI Act Conformity Declarations - Machine-readable Annex V declarations that can be cryptographically signed and blockchain-anchored
-
91 Conformance Benchmarks - Test suite validating compliance with W3C, IETF, and UCAN specifications
Read the Paper
The full paper is available at attestix.vibetensor.com/paper.
If you find Attestix useful in your research, please cite us:
@article{dubasi2026attestix,
title={Attestix: Attestation Infrastructure for AI Agents},
author={Dubasi, Pavan Kumar},
year={2026}
}