npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@arcproof/sdk

v0.1.0

Published

Trust layer for AI agents: make a claim, independently verify it against live data, release real on-chain payment only if it checks out. Bring your own agent (any framework) -- this is the verify+pay primitive, not a fixed agent roster.

Readme

@arcproof/sdk

The trust layer for AI agents: make a claim, independently verify it against live data, release real on-chain payment only if it checks out. Bring your own agent -- any framework, any vertical. This package is the verify+pay primitive; it is not a fixed set of agents.

Extracted from ArcProof's reference implementation (a DeFi protocol-diligence network) once that system was proven working end to end on Arc testnet with real Circle Wallets, a real deployed escrow contract, and real x402-shaped payments. Everything DeFi-specific was left behind; what's left is domain-agnostic.

Why this exists

Payment rails (x402) and agent identity/reputation (ERC-8004-style registries) tell you money moved. They don't tell you the work was correct. This package is the missing verification step: a specialist agent drafts a claim, a deterministic verifier you write independently re-derives the same fact from a canonical source, and payment is a real, on-chain-enforced conditional release based on whether they agree.

lock budget --> gather claims (any agent) --> verify (your rules) --> release / refund
   (real tx)      (bring your own)          (deterministic,           (real tx, enforced
                                              zero LLM judgment)        by a smart contract)

Install

npm install @arcproof/sdk

Quickstart

import { VerifierRegistry, runTrustedJob, escrow, ARC_TESTNET, type WalletCredential } from "@arcproof/sdk";

// 1. Register how to independently check each claim_type your agents produce.
//    Deterministic, zero LLM calls -- this is what keeps a verdict auditable.
const verifiers = new VerifierRegistry();
verifiers.register("apr_rate", async (claim, context) => {
  const trueRate = await lookUpTrueApr(context.loanId); // however you get a canonical value
  const claimed = Number(claim.claim_value);
  const delta = ((claimed - trueRate) / trueRate) * 100;
  return {
    status: Math.abs(delta) <= 2 ? "match" : "mismatch",
    value: trueRate,
    source: "your-canonical-source",
    delta,
    note: `independent true rate: ${trueRate}`,
  };
});

// 2. Deploy (once) a real escrow contract -- or reuse one you already deployed.
const settlerAddress = "0x...";
const contractAddress = await escrow.deployEscrow(ARC_TESTNET, deployerPrivateKey, settlerAddress);

// 3. Run a job: lock -> gather -> verify -> settle/refund. gatherClaims can be
//    anything -- a LangChain.js agent (@arcproof/sdk-langchain), an ElizaOS
//    action (@arcproof/sdk-elizaos), or a plain async function.
const requester: WalletCredential = { kind: "plain", privateKey: requesterPrivateKey };
const settler: WalletCredential = { kind: "plain", privateKey: settlerPrivateKey };

const result = await runTrustedJob(
  { network: ARC_TESTNET, contractAddress, verifiers },
  {
    jobId: "job-123",
    budgetAmount: 0.05,
    requester,
    settler,
    providerAddresses: { "my-agent-v1": "0x..." },
    gatherClaims: async (context) => myAgent.run(context),
    context: { loanId: "loan-001" },
  }
);

console.log(result.overall_verdict, result.total_paid_usdc);

What's real, not simulated

  • Payments: real signed transactions on whatever EVM chain you point NetworkConfig at (defaults ship for Arc testnet, where USDC is the native gas-equivalent currency -- see ARC_TESTNET).
  • Escrow: a real deployed smart contract (escrow.deployEscrow) enforces lock/release/finalize/refund on-chain -- withheld funds simply never leave the contract, not an application-level promise.
  • Circle Developer-Controlled Wallets: pass a { kind: "circle", walletId, circleConfig } credential anywhere a WalletCredential is expected instead of a plain private key -- every payment then signs through Circle's own transaction API. See "Circle Wallets setup" below.
  • Verification: whatever your Verifier functions do. The SDK doesn't call an LLM anywhere in the verification path -- that's the entire point.

Circle Wallets setup (optional, real)

import { circleWallet } from "@arcproof/sdk";

const config = { apiKey: process.env.CIRCLE_API_KEY!, entitySecret: process.env.CIRCLE_ENTITY_SECRET! };
// One-time per Circle account (semi-irreversible):
//   1. generate + register an entity secret (see Circle's docs / the
//      generateEntitySecret / registerEntitySecretCiphertext functions in
//      @circle-fin/developer-controlled-wallets)
//   2. create a wallet set + wallet:
const walletSetId = await circleWallet.createWalletSet(config, "my-app");
const wallet = await circleWallet.createWallet(config, walletSetId, "ARC-TESTNET");
// fund `wallet.address`, then use it as a WalletCredential:
const requester: WalletCredential = { kind: "circle", walletId: wallet.walletId, circleConfig: config };

API surface

  • ClaimSchema / Claim -- a claim is { claim_id, job_id, provider_agent_id, claim_type: string, claim_text, claim_value, provider_source, simulated, verification_* }. claim_type is a plain string you define, not a fixed enum.
  • VerifierRegistry -- .register(claimType, verifier), .verifyClaims(claims, context). A claim with no registered verifier becomes "unverifiable" (never counts toward mismatches or payment) rather than passing or failing.
  • computeJobVerdict / computeProviderPayout / settle -- the payout math: full payout if 0 mismatches, 50% if exactly 1, withheld if 2+; job verdict accept/partial/reject by the same rule at the job level. Pure functions you can call standalone, or settle() which also executes the real release()/finalize() contract calls.
  • hasCheckableClaims(claims) -- guards against the "every provider failed, job silently accepts with money stuck in escrow forever" trap. runTrustedJob already checks this for you.
  • runTrustedJob(config, params) -- the one high-level helper: lock, gather, verify, settle-or-refund.
  • chain -- raw transfer/verifyTransfer/getBalance against any EVM chain.
  • escrow -- deployEscrow/lock/release/finalize/refund/getJob, every function taking a WalletCredential directly (no fixed role system).
  • circleWallet -- thin wrapper over @circle-fin/developer-controlled-wallets.

Adapters

  • @arcproof/sdk-langchain -- wraps a LangChain.js tool-calling agent as a gatherClaims() function, and a createLangChainOrchestrator() that dynamically decides which of several registered specialists a specific request needs (the full orchestrator → specialists → evaluator pattern, not just one wrapped agent).
  • @arcproof/sdk-elizaos -- exposes the same flow as a real ElizaOS Action/Plugin; its gatherClaims can be a single gatherer or a full LangChain orchestrator.

See examples/lending-apr-agent for a complete, non-DeFi worked example (a lending platform's "true APR" + eligibility diligence network, two specialists behind an orchestrator) proving none of this is tied to the reference app's vertical.

License

MIT