arbitova-mcp-server
The arbitova-mcp-server enables AI agents to manage non-custodial USDC escrow payments with integrated AI arbitration on the Base blockchain.
Create Escrow (
arbitova_create_escrow): Lock USDC into the EscrowV1 smart contract, specifying seller, amount, delivery/review deadlines, and averificationURIdefining delivery criteria.Mark Delivered (
arbitova_mark_delivered): Sellers signal completion by submitting a stable public URL for the deliverable — hashed and stored on-chain.Confirm Delivery (
arbitova_confirm_delivery): Buyers confirm satisfactory delivery, releasing funds to the seller (minus a 0.5% protocol fee).Dispute a Delivery (
arbitova_dispute): Buyer or seller opens a dispute citing unmet criteria, triggering AI arbitration to resolve the payment.Cancel Undelivered Escrow (
arbitova_cancel_if_not_delivered): Buyer reclaims a full refund if the seller misses the delivery deadline.Escalate Expired Review (
arbitova_escalate_if_expired): Any address can permissionlessly escalate a stalledDELIVEREDescrow toDISPUTEDafter the review window expires.Get Escrow State (
arbitova_get_escrow): Query on-chain state including buyer/seller info, amount, deadlines, status (CREATED,DELIVERED,RELEASED,DISPUTED,RESOLVED,CANCELLED), and delivery hash.
Funds are held by the smart contract (non-custodially), and the protocol supports direct integration with AI agents (e.g., Claude, LangGraph, CrewAI) for trustless agent-to-agent payment settlement.
Integrates with Circle's USDC stablecoin on Base to facilitate non-custodial escrow payments.
Enables CrewAI agent crews to execute escrow-based transactions with on-chain settlement and arbitration.
Provides tools for building LangGraph-based agent workflows that include escrow creation, delivery, and dispute resolution using Arbitova.
Arbitova
Non-custodial USDC escrow + AI arbitration for agent-to-agent payments on Base.
Two agents lock USDC into a contract, one delivers, the other confirms or disputes, and a neutral AI arbiter resolves. Arbitova never holds the money — the contract does.
No API keys. No registration. No custody. Your Ethereum address is your identity.
Contract:
EscrowV1at0xA8a031bcaD2f840b451c19db8e43CEAF86a088fCon Base Sepolia — verified on Basescan, mainnet launching after auditSpec:
A2A-ESCROW-RFC-v0.1Machine-readable descriptor:
/.well-known/arbitova.jsonPer-case verdict dashboard: arbitova.com/verdicts
Integration paths: arbitova.com/integrate (raw contract / reference SDKs / agent-native)
Reference UI (non-canonical, for humans): arbitova.com/pay
15-minute tutorial:
docs/tutorials/15-min-paid-agent.md
Why this exists
Every A2A / agent-commerce spec in the wild — MCP, Google's A2A, ERC-7683, Coinbase's Agent Commerce — defines how agents talk. None of them define how money moves when the agents don't trust each other.
Arbitova is the missing settlement primitive:
Deterministic state machine.
createEscrow → markDelivered → {confirmDelivery | dispute → resolve | cancel}. No hidden branches, no admin override.No auto-release after timeout. Review windows expire into
DISPUTED, not into seller payout. Silence is safer than a wrong confirmation.Content-hash pinned on-chain. Sellers can't swap the delivery file after the buyer inspects.
Per-case verdict transparency. Every arbiter decision is a signed JSON blob; its
keccak256is stored on-chain. The full verdict history is queryable at/verdicts— no aggregation, no delay.
This is not a marketplace. There is no Arbitova account, no listing fee, no Pro tier. The protocol is the whole product.
Related MCP server: idios
Quick start — Node.js SDK
npm install @arbitova/sdk ethersimport { Arbitova } from '@arbitova/sdk';
const buyer = await Arbitova.fromPrivateKey({ privateKey: process.env.BUYER_PK });
const { escrowId, txHash } = await buyer.createEscrow({
seller: process.env.SELLER_ADDRESS,
amount: '5.00',
deliveryHours: 24,
reviewHours: 24,
verificationURI: 'https://example.com/spec.json',
});
console.log(`Escrow #${escrowId} locked — ${buyer.explorerTx(txHash)}`);Seller-side, arbiter-side, browser wallet integration: see packages/sdk-js/README.md.
Quick start — Python SDK
pip install "arbitova[path_b]"from arbitova import path_b
result = path_b.arbitova_create_escrow(
seller="0x...",
amount=5.00,
verification_uri="https://example.com/spec.json",
)
print(result)Quick start — Claude / any MCP client
{
"mcpServers": {
"arbitova": {
"command": "npx",
"args": ["-y", "@arbitova/mcp-server"],
"env": {
"ARBITOVA_RPC_URL": "https://sepolia.base.org",
"ARBITOVA_ESCROW_ADDRESS": "0xA8a031bcaD2f840b451c19db8e43CEAF86a088fC",
"ARBITOVA_USDC_ADDRESS": "0x036CbD53842c5426634e7929541eC2318f3dCF7e",
"ARBITOVA_AGENT_PRIVATE_KEY": "0x..."
}
}
}
}Six tools: arbitova_create_escrow, arbitova_mark_delivered, arbitova_confirm_delivery, arbitova_dispute, arbitova_cancel_if_not_delivered, arbitova_get_escrow. All sign locally via ethers v6. Your private key never leaves the process.
Omit ARBITOVA_AGENT_PRIVATE_KEY for read-only introspection mode (useful for observability).
Lifecycle
┌──────────────────┐
│ CREATED │ buyer locked USDC
└────────┬─────────┘
│
▼ seller.markDelivered()
┌──────────────────┐
│ DELIVERED │ deliveryHash on-chain
└────────┬─────────┘
│
buyer.confirmDelivery()│ │ buyer.dispute()
│ │ or seller.dispute()
▼ ▼
┌─────────────┐ ┌──────────┐
│ RELEASED │ │ DISPUTED │ waiting for arbiter
└─────────────┘ └────┬─────┘
│ arbiter.resolve(bps split + verdictHash)
▼
┌──────────┐
│ RESOLVED │
└──────────┘Two terminal states not drawn: CANCELLED (buyer calls cancelIfNotDelivered after delivery window) and auto-escalation into DISPUTED if the review window expires without confirmation.
Framework reference agents
Three end-to-end A2A demos on Base Sepolia with a live AI arbiter:
Claude Agent SDK — in-process MCP tools
LangGraph — ReAct agent, buyer + seller + arbiter
CrewAI — Agent + Task + Crew
Each demo runs the full CREATED → DELIVERED → CONFIRMED (or DISPUTED → RESOLVED) flow with real on-chain transactions.
Packages
Package | Purpose |
Node.js / browser SDK ( | |
Python SDK, install with | |
MCP server (6 on-chain tools) for Claude Desktop, Claude Code, any MCP client |
Each ships the same six-entrypoint surface so an agent using the Python SDK can settle with an agent using the MCP server — they're hitting the same contract.
Fees
When | Fee | Paid by |
| 0.5% | deducted from seller payout |
Arbiter resolves a dispute | 2% | split per arbiter verdict |
Fees accrue in the contract. The protocol runs on them; there is no subscription.
Networks
Network | Status | Contract |
Base Sepolia | live, real Circle USDC |
|
Base mainnet | pending audit + multisig arbiter | TBA |
Watch the Dev Log for mainnet launch.
Legacy (Path A)
v2.x of the SDKs and v3.4.0 of the MCP server were a custodial HTTP client against api.arbitova.com. That architecture had four structural problems (DB-vs-onchain drift, custody wallet gas, single ADMIN_KEY, single WALLET_ENCRYPTION_KEY point of failure) and was deprecated in favor of Path B — the non-custodial on-chain design described above.
Migration for SDK users:
sdk/MIGRATION_PATH_A_TO_B.mdMigration for MCP users:
mcp-server/MIGRATION.md
Old packages remain on npm/PyPI but are deprecated.
License
MIT
Available Tools
7 toolsarbitova_cancel_if_not_deliveredA
Buyer cancels an escrow after the delivery deadline has passed and the seller has not marked delivery. Full USDC refund to buyer. Only callable by the buyer, only after deliveryDeadline has elapsed, and only when escrow is still in CREATED state. Call arbitova_get_escrow first to verify the deadline has passed and status is CREATED before calling this.
Inputs: escrowId (required).
arbitova_confirm_deliveryA
Confirm that the seller has delivered and the output meets ALL criteria in the escrow's verificationURI. ONLY call after you have: (1) fetched the delivery payload, (2) fetched and parsed the verificationURI criteria document, (3) compared the payload against EVERY criterion, (4) are fully confident every criterion is satisfied. If anything is missing, wrong, ambiguous, incomplete, or the seller did not deliver the agreed output — DO NOT call this. Call arbitova_dispute instead with a specific reason citing the failed criterion. Silence is safer than a wrong confirmation: if you do not confirm within the review window, funds auto-escalate to AI arbitration, which protects the buyer by independently evaluating delivery.
Inputs: escrowId (required).
arbitova_create_escrowA
Buyer locks USDC into the Arbitova EscrowV1 smart contract. Calls USDC.approve() then createEscrow() on-chain. Requires USDC balance >= amount. deliveryWindowHours = how long the seller has to deliver (default 24). reviewWindowHours = how long the buyer has to verify after delivery is marked (default 24). verificationURI must point to a publicly fetchable JSON document listing every criterion the delivery will be checked against. If the review window expires without confirmation or dispute, funds auto-escalate to arbitration. Silence protects the buyer — you do NOT need to confirm promptly.
Inputs: seller (required), amount (required), verificationURI (required), deliveryWindowHours, +1 more.
arbitova_disputeA
Open a dispute when delivery is missing, incorrect, incomplete, or violates the verification criteria. Disputes trigger AI arbitration which independently evaluates the delivery and can split funds fairly. Call this whenever arbitova_confirm_delivery feels uncertain — when in doubt, dispute. The reason field MUST cite specific criteria from verificationURI that were not met, e.g. "Delivery at [URI] is missing criterion 3: word count is 400 but spec requires >= 1000." Either the buyer OR the seller can open a dispute.
Inputs: escrowId (required), reason (required).
arbitova_escalate_if_expiredA
Permissionlessly escalate a DELIVERED escrow to DISPUTED after the review window has expired. Any address can call this — buyer, seller, or an independent watcher. Use when the buyer went silent past reviewDeadline: this unblocks the seller and routes the escrow into arbitration instead of leaving funds stuck. Only valid when status is DELIVERED and block.timestamp > reviewDeadline. Call arbitova_get_escrow first to verify status and deadline before calling this.
Inputs: escrowId (required).
arbitova_get_escrowA
Fetch the current on-chain state of an escrow: buyer, seller, amount, deadlines, status, verificationURI, and deliveryHash. Use this to check whether delivery has been marked before fetching the payload, and to verify the reviewDeadline before deciding to confirm or dispute. Status values: CREATED (awaiting delivery), DELIVERED (seller marked done, review window open), RELEASED (funds released to seller), DISPUTED (in arbitration), RESOLVED (arbiter resolved), CANCELLED.
Inputs: escrowId (required).
arbitova_mark_deliveredA
Called by the SELLER after completing the work and uploading the deliverable to a stable URL. deliveryPayloadURI must point to the completed, publicly fetchable deliverable — the contract hashes this URI on-chain. Do NOT call before work is actually done. Do NOT use a temporary or expiring URL. The arbiter will fetch this URI during dispute resolution — a broken or mismatched URL counts as non-delivery. Before calling: complete all work to spec, upload to a stable URL (IPFS or persistent storage), and verify the content matches every criterion in verificationURI.
Inputs: escrowId (required), deliveryPayloadURI (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 7 tools
Each tool targets a unique action in the escrow lifecycle (create, get, mark delivered, confirm, cancel, dispute, escalate). No two tools have overlapping purposes, and descriptions clarify exactly when each should be used.
All tools follow the pattern 'arbitova_<verb>_<condition>', using snake_case consistently. The names are descriptive and predictable, making it easy to infer the tool's action from its name.
7 tools is an ideal size for an escrow management server. Each tool serves a necessary function in the lifecycle without unnecessary redundancy or bloat.
The tool set covers the full escrow lifecycle: creation, status retrieval, delivery marking, confirmation/cancellation, dispute opening, and escalation. There are no obvious gaps that would hinder an agent from managing an escrow end-to-end.
Related MCP Connectors
Agent-to-agent escrow on Base. Post quests with ETH/USDC bounties and settle on-chain.
Escrow middleman for any deal between people or AI agents: USDC held in an audited contract.
Escrow protection for agent payments on Base — USDC held in smart contract until job completion.
Escrow protection for agent payments on Base — USDC held in smart contract until job completion.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceSettlement rails for AI labor — USDC escrow on Base Mainnet, 1% protocol fee, designed for autonomous agents. 10 MCP tools covering the full escrow lifecycle: * Quoting calldata for create-intent, submit-proof, release-funds (broadcast gated) * Single-call x402 payment binding (replaces the 5-step x402 dance with one HMAC-signed POST) * Server-side reputation from on-chain event scan * LiMIT
- AlicenseAqualityCmaintenancePrivate escrow for AI agent work on Beam mainnet an agent locks payment, the worker locks collateral, and delivery settles on hash match or review, with M of N arbitrator voting and slashable worker bonds as the dispute backstop. 22 tools cover the full contract lifecycle, and dispute voting is deliberately not an agent tool, so an agent can never rule in its own favour.26MIT
- AlicenseNot gradedqualityDmaintenanceMulti-party escrow and settlement for autonomous agent transactions with x402 micropayments and dispute resolution on Base, Ethereum, or Solana.MIT
- AlicenseAqualityBmaintenanceZero-key on-chain intelligence for Base, Ethereum, and Solana. Instant bytecode decompilation, honeypot detection, EIP-1967 proxy resolution, calldata/revert decoding, and pre-flight transaction simulation. No API keys or wallet connection required.7132MIT