Agentic payments on Arc, and where the infrastructure actually stands

Circle has shipped real, testnet-live building blocks for agent-to-agent payments on Arc — an ERC-8004-based agent identity registry, ERC-8183 job-and-escrow contracts, and x402-based micropayments, all with working tutorials as of 2 September 2026. That is genuine infrastructure, not just an announcement. But it is testnet-only, pre-mainnet, and built by Circle itself rather than by named third-party projects. Tempo, a competing stablecoin chain from Stripe and Paradigm, launched its mainnet on 18 March 2026 already bundled with a machine-payments protocol — meaning a direct competitor has agentic payments live in production while Arc's equivalent is documented, testable, and still pre-mainnet.

What "agentic payments" means as a category

The pitch is autonomous AI agents transacting with each other or with services directly — an agent paying a per-call fee for an API, an agent hiring another agent for a task and paying on delivery — without a human approving each transaction. It needs three things at minimum: a way for an agent to have a verifiable onchain identity, a way to structure a job with escrow and settlement conditions, and a way to move very small payments cheaply enough that micropayments make economic sense. Whether a chain has *shipped* this, versus merely discussed it, comes down to whether those three pieces exist as working, testable infrastructure.

What Circle has actually published for Arc

Circle's own documentation describes a specific, working set of components under an "agentic economy" heading:

• ERC-8004 identity — an onchain agent identity, reputation, and validation framework. Circle's tutorial deploys three live testnet contracts (an IdentityRegistry, ReputationRegistry, and ValidationRegistry) and walks through registering an agent against them — this is a working testnet flow, not a whitepaper description.
• ERC-8183 jobs — a job-lifecycle standard covering job creation, USDC escrow funding, deliverable submission, evaluation, and automated settlement between agents.
• x402 micropayments — support for the x402 protocol, described in Circle's own material as enabling an agent to "pay for premium API endpoints in USDC fractions" via what Circle calls Nanopayments.
• Sample applications — Circle's own reference implementations, Arc Escrow (AI-validated freelance work with USDC settlement) and Arc Nanopayments (autonomous agent payments for paywalled API access), both described as open-source examples rather than maintained third-party products.
• Stated infrastructure advantages — Circle's documentation frames Arc's sub-second finality and USDC-native settlement (no volatile gas token) as suited to agent transaction volumes, with compliance screening for agent payment flows via Elliptic and TRM Labs.

This is a materially more developed starting point than "Arc has announced interest in AI agents" — the identity and job-settlement contracts are deployed and documented with working tutorials. What it is not, as of 2 Sep 2026, is a mainnet-live system, and it is not (yet) a layer with named third-party projects building agent-payment products on top of it that Built on Arc could independently confirm.

The competitive comparison: Tempo already shipped this on mainnet

Tempo — the stablecoin L1 built by Stripe and Paradigm, which raised a $500m Series A at a $5bn valuation — launched its mainnet on 18 March 2026, roughly six months ahead of Arc's 16 September 2026 mainnet, and did so already bundled with a machine-payments protocol for agent-driven transactions, described on Tempo's own site as "agentic commerce" enabling low-cost, instant payments for agents to autonomously execute transactions. Built on Arc could not source a specific named protocol brand for Tempo's agent-payments feature beyond that framing, but the material point stands regardless of naming: Tempo shipped an agent-payments capability in production before Arc's equivalent infrastructure has even reached its own mainnet. Combined with Stripe's existing merchant distribution, that makes agentic payments a live competitive axis between the two chains right now, not a future one.

The honest read on where Arc stands

Arc's agent-payment infrastructure is real and testnet-live — that's a stronger position than "announced only." But relative to Tempo, it is not yet shipped in the sense that matters most for a builder deciding where to deploy an agent-payments product today: production traffic, real settlement, mainnet economics. A team evaluating Arc for an agentic-payments use case should treat ERC-8004/ERC-8183/x402 as genuinely usable testnet infrastructure to build and test against now, while recognizing that going live means waiting on Arc's own mainnet timeline, and that a direct competitor is already collecting production usage and merchant integrations in the meantime.

Sources: - Arc agentic economy overview - Arc: register your first AI agent (ERC-8004 tutorial) - Arc sample applications reference - Tempo - Arc compliance vendors reference

*Built on Arc is an independent directory. Arc is a Circle product; we are not affiliated with, endorsed by, or operated by Circle.*