IRIS

IRIS solves the crisis of anonymity and accountability that currently prevents the AI agent economy from scaling. When you hire a human auditor, you check LinkedIn, references, and degrees. When you hire an anonymous AI bot, you have nothing.

IRIS provides the three pillars of trust necessary for safe commerce:

1. The Uncorruptible Passport (Identity) IRIS gives every autonomous agent a permanent digital identity tied to a unique address.

The Problem It Solves: It prevents Sybil attacks, where one bad agent creates 10,000 fake identities to flood the market or evade accountability.

The Human Equivalent: Like having a passport and a Social Security Number that proves you are a unique individual. Once registered on IRIS, that agent is permanently on the record.

1. The All-Seeing Resume (Reputation) IRIS creates an immutable, uncensorable service history for every agent, based on actual outcomes.

The Problem It Solves: It stops agents from deleting bad reviews or faking their service history. If an agent performs a task, the client's feedback (score and detailed review) is permanently recorded and instantly visible.

The Human Equivalent: Like a credit score or an audit report that cannot be altered, even by the agent itself. Recruiters (clients) can see the average score and the hash (proof) of the detailed, off-chain review before committing any funds.

1. The Proactive Security Guard (The Active Eye) IRIS ensures safety by automatically protecting the marketplace from verified malicious or incompetent agents.

The Problem It Solves: In a global, 24/7 market, you can't wait for a human to ban a bad actor. IRIS uses automation (Chainlink) to constantly watch the reputation history.

The Human Equivalent: If the agent’s reliability score falls below the legal minimum, the system automatically revokes its operating license and flags it as Suspended. This happens instantly, without human approval, protecting everyone else in the system

IRIS is built as a fully modular, end-to-end agent lifecycle framework that combines on-chain identity, decentralized storage, and automated reputation computation. At its core, I built a TypeScript SDK exposing three modules—Identity, Storage, and Registry—designed around ERC-8004’s identity packet model for agents.

The Identity module signs agent metadata using ECDSA and generates deterministic agent IDs. The Storage module integrates directly with 0G’s decentralized storage network: metadata packets are chunked, uploaded through the 0G Indexer, and resolved using verifiable Merkle-based data hashes. I had to patch and work around parts of their uploader logic—for example when the storage market contract returned empty selectors—to get fully consistent uploads. That hack saved the build.

The Registry smart contract is deployed on Base Sepolia, chosen for its finality speed and tight EVM compatibility with 0G and Chainlink. It stores agent identity roots, metadata hashes, and trust/reputation scores. The most interesting part is the Chainlink Automation integration: whenever an agent receives a new review, the contract emits a ReviewSubmitted log. A Chainlink job listens for that log, pulls the latest metadata packet from 0G using Any API, recomputes the agent’s score off-chain, and automatically updates it on-chain. That makes IRIS a self-maintaining, trust-minimized agent ecosystem.

To make this accessible, I also built a frontend UI that visually walks users through registering an AI agent—either fully on-chain (Base + 0G) or as a Web2-only registration for prototyping. The dashboard shows the full lifecycle: identity packet creation, 0G upload status, transaction confirmations, and real-time score updates streamed from Chainlink events. This UI makes the whole system feel alive and understandable even for non-technical users.