Skip to main content
AgentLedger is the missing trust infrastructure for autonomous AI agents. Over 104,000 AI agents are now registered across 15+ incompatible registries — yet no standard exists for an agent to verify whether a service is safe to call. AgentLedger fills that gap with three integrated components: a universal Manifest Registry for service discovery, a blockchain-anchored Trust Ledger for immutable attestation, and an Audit Chain for liability-grade action logs.

Introduction

Understand what AgentLedger does and why trust infrastructure is the critical missing layer in the agent web.

The Problem

Explore the discovery gap, trust vacuum, and fragmentation crisis driving the need for AgentLedger.

Architecture Overview

See how the three components — Manifest Registry, Trust Ledger, and Audit Chain — work together.

Manifest Registry

Learn how services register capabilities and how agents discover and query them.

Trust Ledger

Understand blockchain-anchored attestations and how trust scores are computed.

Audit Chain

See how every agent action is logged in a tamper-proof, liability-grade chain of custody.

​
The Three Components

AgentLedger addresses the trust and discovery gaps in today’s agent protocol stack through three integrated components:
1

Manifest Registry

Services publish a structured manifest at /.well-known/agent-manifest.json declaring their capabilities, pricing, data requirements, and SLAs. Agents query the registry by capability ontology tag, natural language, or ambient discovery to find verified services before interacting with them.
2

Trust Ledger

Attestation events, revocation records, and version history are anchored on-chain — making trust claims independently verifiable by any agent without relying on AgentLedger directly. No single party, including AgentLedger, can retroactively alter a service’s trust history.
3

Audit Chain

Every consequential agent action — bookings, payments, record access — is captured in a PII-stripped, cryptographically linked audit record. This enables dispute resolution, liability attribution, and EU AI Act compliance without exposing sensitive user context.

​
Reference

Capability Ontology

The shared vocabulary — 5 root domains, 65 capability tags — through which services declare what they do.

Threat Model

Six primary attack vectors against the trust layer and the mitigations built into each component.

Roadmap

The four-phase path from open standard to infrastructure status.