IRSB Protocol

  • Name: IRSB Protocol
  • URL: https://github.com/intent-solutions-io/irsb-protocol
  • Category: AI-agent guardrails / intent-solver accountability / delegated-wallet policy enforcement / receipt-and-bond infrastructure
  • Summary: IRSB Protocol is worth cataloging not as another generic agent wallet or intent security package, but as a lower-layer guardrail and recourse stack for machine execution. Its strongest primary materials separate EIP-7702-style delegated wallet execution, onchain caveat enforcers, cryptographic receipts, bonded solver admission, challenge-window disputes, ERC-7683 intent references, optional x402 payment accountability, and ERC-8004 reputation export into distinct comparison-ready layers. That makes IRSB useful for the corpus because it exposes where practical control actually sits when an agent or solver is allowed to act: in the delegate contract, the policy-enforcer set, the receipt format, the watchtower/dispute path, the bond and slashing rules, and the offchain storage and privacy model around evidence.
  • What it does:
    • Uses a WalletDelegate plus enforcer modules to constrain delegated execution by spend caps, time windows, allowed targets, allowed methods, and replay-protection nonces
    • Produces cryptographic receipts for agent or solver actions and routes them through an onchain IntentReceiptHub with challenge windows and finalization
    • Requires bonded registration so solvers or execution providers can be slashed when challenged receipts prove fault
    • Connects receipts to ERC-7683-style intent flows and treats the same machinery as usable for both AI-agent actions and intent-solver execution
    • Adds a privacy architecture where commitments and pointers live onchain while fuller payloads and evidence remain offchain, optionally access-gated or Lit-encrypted
    • Publishes an x402 integration path where HTTP-payment-backed API calls can return IRSB receipts that later support disputes, reputation, and service-accountability checks
    • Frames ERC-8004 as the downstream reputation layer, with IRSB acting as a validation provider that turns execution history into portable trust signals
  • Key claims:
    • IRSB clears the intake bar because it makes agent execution guardrails legible as a reusable protocol surface instead of hiding them inside one wallet SDK, one agent framework, or one intent marketplace.
    • The most important decomposition is policy versus proof versus recourse. Enforcers decide whether an action may execute, receipts commit what allegedly happened, and bonds plus disputes decide what happens if that execution is contested. Many products collapse those layers together.
    • IRSB is analytically useful because it treats AI agents and DeFi solvers as the same broad problem: delegated machine execution with value at risk. That makes it a bridge project between agent-wallet tooling and intent-settlement infrastructure.
    • The x402 integration docs are especially useful because they turn paid API access into a comparison-ready accountability layer: payment proof, service result, receipt generation, optional dual attestation, onchain posting, and later dispute/finalization are all kept explicit.
    • The privacy design matters because the protocol does not simply dump execution details onchain. It explicitly separates commitments, CIDs, and access conditions from the underlying request/response payloads, which is a more reusable pattern than generic audit trail rhetoric.
    • The main operational caveat is that IRSB still depends on offchain monitoring, evidence availability, and timely dispute action. A one-hour challenge window, offchain ciphertext pointers, and optional client co-signing mean the recourse model is only as strong as the surrounding watchtower, storage, and participant operations.
    • The repo presents meaningful but still first-party maturity signals: extensive tests, Sepolia deployments, and concrete package boundaries are all useful, but they should still be read as project claims rather than independent audits.
    • This entry belongs in the active corpus because it separates delegated execution policy, receipt attestations, collateralized recourse, payment-linked accountability, and reputation export into reusable comparison layers that would be lost if it were treated only as an AI agent safety repo.
  • Whitepaper: No standalone whitepaper surfaced in this pass. The strongest materials were the main repository README plus the privacy and x402 integration docs collected in ../whitepapers/irsb-protocol-primary-sources-2026-05-13.md.
  • Sources:
  • Last reviewed: 2026-05-13 UTC