Make every agent payment decision portable and replayable.

Your wallet or control provider enforces the decision. Open Agent Spend Guard records which policy and facts produced it, so another operator can reproduce the result across providers.

example-agent / policy draftIllustrative
Maximum per request$500
Human approvalAbove $200
Allowed tokensUSDC, ETH
Allowed chainsBase, Ethereum
Infrastructure reserveApproved
Vendor execution requestApproval
Trading allocationBlocked

Open Agent Spend Guard v0.1.2 is ready to test

The first open release is a real TypeScript package and CLI, not only a browser mockup. It evaluates versioned policies with atomic-integer arithmetic, normalizes x402 v2 and bounded Safe requests, and exports reproducible RFC 8785 evidence packages.

Release proofShipped in v0.1.2
Versioned artifactsPolicy, request, decision, and complete evidence-package schemas
Adaptersx402 v2 exact payments and read-only Safe native transfers
Verification102 automated tests, golden hashes, strict typecheck, green public CI, tagged-install proof, and a reproducible Base Sepolia x402 challenge capture
Security boundaryNo keys, signing, custody, settlement, Safe calldata, delegate calls, or gas refunds

Inspect the Base Sepolia x402 proof

The public walkthrough uses the official x402 Express middleware and public facilitator to return a real unauthenticated HTTP 402 challenge for Base Sepolia USDC. OASG normalizes that challenge and exports replayable allow, approval-required, and block evidence packages.

The walkthrough does not use a wallet key or claim a funded settlement. Its receiver is an inert demonstration address. That keeps the proof reproducible without creating custody or production-fund risk.

Enforcement is improving. Decision evidence is still fragmented.

Wallet providers and payment-control products increasingly ship native limits, ledgers, modules, and guards. Those systems should remain authoritative. Their policy meanings and audit records do not automatically travel when an agent touches multiple wallets, chains, facilitators, and providers.

Open Agent Spend Guard is the neutral evidence layer between them. It normalizes intent, produces a deterministic reference decision, and exports a replayable record without pretending to replace runtime enforcement.

Start with a paid policy and evidence workshop

The initial commercial entry point is a fixed workshop, not another wallet or a risky enforcement deployment.

  • Agent, wallet, provider, and payment-flow inventory
  • Budget hierarchy and authority model
  • Token, chain, counterparty, amount, and timing rules
  • Human approval and exception evidence
  • Replay, reconciliation, and audit requirements
  • Provider control mapping and evidence-emitter roadmap

Workshop price: $3,500 paid upfront for one agent, one provider, and one operating environment. Multi-provider or production work is separately scoped.

Design-partner product scope

CapabilityWhat it doesInitial mode
System inventoryMaps agents, owners, wallets, providers, and environmentsWorkshop artifact
Policy schemaRepresents versioned amount, asset, chain, counterparty, purpose, and approval rulesOpen v0.1 schema
Reference evaluatorProduces deterministic allow, block, or approval-required decisionsOpen v0.1 package
Evidence packageBinds the policy, request, reason, decision, timestamp, and canonical hashesReplayable JSON
NormalizersMaps provider-native requests into the shared evidence modelx402 and Safe in v0.1
Emitter mappingShows an existing wallet or control product how to emit portable evidenceDesign-partner scope

Connector order

  1. Open schemas, evaluator, evidence package, and CLI - shipped in v0.1
  2. x402 v2 and Safe native-transfer normalizers - shipped in v0.1
  3. Base Sepolia x402 challenge and evidence walkthrough - shipped publicly
  4. Three qualified design-partner evaluations and an independent review - recruiting now
  5. Evidence emitters for existing wallet and enforcement providers

Design-partner terms

No-meeting replay: send one sanitized synthetic or testnet x402 or Safe request. We will return the normalized request, three policy outcomes, a replayable evidence package, and a concise issue list. No production keys, funds, screen share, or live meeting are required.

A bounded evidence-emitter prototype begins at $10,000 for one real payment flow, one existing wallet or control provider, a test environment, and explicit responsibility boundaries. Production integration begins at $25,000 after a successful workshop or prototype. OASG does not need custody or signing authority for either scope.

Public-good track

The provider-neutral schemas, reference evaluator, evidence format, and selected read-only normalizers are suitable for ecosystem grant funding. Commercial workshops, evidence-emitter mappings, and client-specific integrations remain separately scoped.

Make the next alert easier to own.

Start with a fixed-scope readiness sprint. Keep the tools you already trust.