avellum.ts

 1  |

Protocol
documentation.

Technical overview of the Avellum trust layer, verifier economics, and API reference.

MODULEDOCUMENTATION

VERSIONV2.0

1. Executive Summary

SYS_DOC_V2

The A2A (Agent-to-Agent) economy is growing rapidly, with AI agents transacting billions without human oversight. However, there's no standardized way to verify which agents are trustworthy before interacting.

Avellum solves this by providing a decentralized trust layer that indexes agents across x402, MCP, and A2A protocols, allowing token-weighted verification to establish a definitive “Trust Score” for every autonomous actor on the network.

2. The Problem

ERR_01No unified trust standard for AI agents operating across disparate chains and runtimes.

ERR_02Agents operate in siloed protocols (x402 vs MCP vs A2A) making reputation fragmented.

ERR_03No verifiable way to assess reliability before initiating value transfer.

ERR_04Centralized rating systems create single points of failure and censorship risks.

3. How Avellum Works

x402
Protocol

MCP
Registry

A2A
Protocol

INGESTION

AVELLUM INDEXER

Normalization

Verifier A

Verifier B

Verifier C

TRUST SCORE API

Public Access

Avellum operates as a multi-protocol indexer. It listens to events on major agent standards (x402, MCP, A2A), normalizes the identity data, and passes it to a decentralized network of Verifiers. Verifiers stake $AVLM tokens to rate agents based on performance, uptime, and security audits.

4. Trust Score Algorithm

trust_score = Σ(rating × token_weight) / Σ(token_weight)

// Weighted average based on verifier stake

The trust score is not a simple average. It is stake-weighted. A verifier with 100,000 $AVLM tokens has 10x the voting power of a verifier with 10,000 tokens. This Sybil-resistance mechanism ensures that those with the most to lose have the most say in the network's reputation layer.

5. Verifier Economics

Revenue Model

100% to Verifiers

Distribution

Proportional to Activity

The Avellum protocol takes 0% fees. All revenue generated from API keys (consumed by developers querying trust scores) is distributed directly to the Verifier pool. Active verifiers who consistently rate new agents earn the highest yield.

6. API Reference

GET/api/score/[agent_address]

{
  "address": "0x7a2...9f1",
  "trust_score": 94.5,
  "confidence": "HIGH",
  "signals": {
    "uptime": "99.9%",
    "avg_latency": "240ms",
    "verified_txns": 14205
  },
  "last_updated": "2024-03-15T10:30:00Z"
}

7. Supported Protocols

x402 Protocol

Payment-enabled autonomous agents standard.

x402

MCP Registry

Anthropic's Model Context Protocol.

MCP

A2A Standard

Google's Agent-to-Agent communication layer.

A2A

8. Token Utility

$AVLM is the native utility token of the network. It is used for:

Verification Weight

More tokens = higher influence on trust scores.

Stake-to-Earn

Staked tokens earn a share of API protocol revenue.

Governance

Voting on protocol upgrades and new registry integrations.

9. FAQ

How fast do trust scores update?

Scores are updated in near real-time (every ~5 minutes) as new verifier signatures are aggregated on-chain.

Can bad actors game the system?

Slashing conditions exist. If a verifier approves a malicious agent that is later proven fraudulent, a portion of their staked $AVLM is slashed.

Is there a developer API/SDK?

Yes, check the API Reference section above or visit our GitHub for the Node.js and Python SDKs.

Protocoldocumentation.