The accountability layer for AI agents
Every agent action. Receipted. Bonded. Monitored.
AI agents are getting wallet access. IRSB ensures they can't overspend, can't call unauthorized contracts, and can't act without cryptographic proof. On-chain policy enforcement with automated dispute resolution and portable reputation.
The Problem
AI agents are executing transactions with real money and no guardrails.
No Spend Limits
An agent with wallet access can drain funds. There are no on-chain controls limiting what it can spend, which contracts it can call, or when it can act.
$0
recovery if your agent overspends
No Proof of Work
Solvers and agents claim they executed correctly. There is no standardized receipt, no cryptographic proof, and no way to verify after the fact.
0
on-chain receipts in existing systems
No Consequences
Bad actors walk away. No bonds at risk, no automated dispute process, no reputation impact. Honor systems do not scale.
0%
of solver reputation is portable today
What IRSB Does
Three layers of protection, enforced at the blockchain level.
Policy Enforcement
Agents delegate their wallet to smart contract code that enforces spend limits, time windows, approved contracts, and approved functions. Violations are impossible — they revert at the EVM level.
Powered by EIP-7702 + 5 caveat enforcers
Cryptographic Receipts
Every execution produces an on-chain receipt with signed evidence. V2 receipts include dual attestation — both the solver and client sign, creating non-repudiable proof of what happened.
EIP-712 signed, on-chain, immutable
Economic Accountability
Solvers stake bonds (min 0.1 ETH) that get slashed for violations. 80% goes to the affected user. A watchtower monitors receipts and files disputes automatically. Three strikes means permanent ban.
Bonded, slashable, automated monitoring
The Intent Lifecycle
From intent to accountability in five steps.
Submit Intent
User expresses what they want done
Solver Executes
A bonded solver picks up and fills the intent
Receipt Posted
Cryptographic proof of execution posted on-chain
Challenge Window
1 hour for anyone to dispute with evidence
Finalize
No dispute = receipt finalizes, reputation updates
Who Uses This
AI Agent Frameworks
AgentKit, ElizaOS, Olas, Virtuals
Agents with wallet access need spend limits, contract whitelists, and audit trails. IRSB enforces these at the EVM level — not in application code.
DeFi Intent Protocols
Across, CoW, UniswapX, 1inch
Users submit swap intents but have no proof solvers executed correctly. IRSB provides standardized receipts, bonded accountability, and portable solver reputation.
Work Coordination Platforms
Bounty platforms, agent marketplaces
When agents complete work for pay, cryptographic receipts replace trust. Escrow locks on claim, releases on verified completion, slashes on violation.
API Payment Verification
Paid APIs, compute marketplaces
x402 HTTP payment integration. When a paid API fulfills a request, a receipt proves service delivery on-chain. No more "did I get what I paid for?" ambiguity.
The Ecosystem
Four systems working together. Each solves a specific accountability gap.
No Proof or Consequences
Protocol
On-chain contracts for receipts, bonds, disputes, and escrow. 3 verified contracts, 308 tests.
No Execution Engine
Solver
Execute intents, produce evidence, submit receipts. Pluggable job types with CLI and HTTP interfaces.
No Independent Monitoring
Watchtower
Monitor receipts, detect violations, file disputes. Rule engine with configurable violation detection.
No On-Chain Policy
Delegation (EIP-7702)
Cloud KMS signing with on-chain policy enforcement via EIP-7702 WalletDelegate and 5 caveat enforcers.
Built, Tested, Deployed
Not a whitepaper. Working infrastructure on Ethereum Sepolia.
11
Verified Contracts
on Sepolia
1,000+
Passing Tests
552 Foundry + 457 TS
6
EIP Standards
integrated
5
CI Workflows
automated
Standards-Based Architecture
Built on emerging Ethereum standards. Each one solves a specific problem.
A standard way to express "what I want done" across chains. IRSB receipts reference these intents.
Lets an agent's wallet delegate to smart contract code that enforces rules. The key innovation for agent guardrails.
Allows batching multiple delegated transactions atomically. Solvers and facilitators use this for efficient execution.
A standard way for apps to request execution permissions from wallets. The user-facing side of delegation.
On-chain identity and reputation for agents. IRSB publishes validation signals that feed into an agent's permanent record.
Pay-per-request for APIs. IRSB adds accountability — when you pay for an API call, a receipt proves you got what you paid for.
Where to Start
Building an AI agent with wallet access?
We are working with agent frameworks and DeFi protocols to integrate on-chain guardrails. Currently deployed on Sepolia testnet.
IRSB is experimental software deployed on Ethereum Sepolia testnet. It has 1,000+ passing tests including fuzz tests but has not yet undergone a formal security audit. Not for production use with real funds. Licensed under BUSL-1.1 (converts to MIT on 2029-02-17).