Introduction

ATLAST Protocol (Agent Trust Layer, Accountability Standards & Transactions) is the trust infrastructure for the AI agent economy.

What Problem Does It Solve?

AI agents are increasingly autonomous — writing code, making decisions, managing workflows. But:

• How do you know what an agent actually did? Not what it claims, but what it provably did.
• How do you verify an agent's work without re-running everything?
• How do you compare agents objectively, based on evidence rather than marketing?

ATLAST Protocol answers all three questions with the Evidence Chain Protocol (ECP) — a cryptographic audit trail for every agent action.

How It Works

Agent does work → ECP records input/output hashes → Links into chain → Signs with Ed25519
                                                                              ↓
                                                          Batch → Merkle Root → EAS on-chain

Think of it as a dashcam for AI agents: it doesn't judge what the agent does, but it proves exactly what happened.

Three-Layer Integration

Layer	Effort	How
Layer 0	Zero code	atlast run python my_agent.py — transparent proxy
Layer 1	5 lines	from atlast_ecp import wrap — SDK wrapper
Layer 2	10-20 lines	LangChain / CrewAI / AutoGen adapters

Key Properties

• Privacy: Content never leaves your device. Only SHA-256 hashes are transmitted.
• Tamper-proof: Chain hashes + Ed25519 signatures. Modifying any record breaks the chain.
• Verifiable: Anyone can verify a record's authenticity without the original content.
• Fail-Open: Recording failures never affect agent operation.
• Open Standard: MIT license, designed for ecosystem adoption.