The wallet that
lets AI agents buy.
Anywhere.
One API that completes checkout at any store — no merchant integration, no funds held, with the cardholder’s signed permission.
Thru
Thru
AI agents can find anything.
They can buy almost nothing.
We finish the purchase. An agent hands Agnic the cart; we complete checkout on the merchant’s existing storefront — no integration on their side, no funds held on ours — and every order carries the cardholder’s signed permission.
Discovery ends in a handoff.
AI names the perfect product and the right store in seconds — then hands the human a link. Between the recommendation and the checkout form, the order dies.
Millions of stores will never integrate.
Agent-checkout protocols only work where merchants adopt them — a handful have. The long tail of shops, restaurants, and booking pages is not adding an API for agents.
Nobody can prove the buyer said yes.
When software spends a person’s money, the merchant and the bank need proof of permission. A card number pasted into a bot proves nothing. A signed, scoped, revocable mandate does.
When agents act,
both sides have proof.
A new way to buy. A new way to be found. The user says it in plain language, the merchant receives a cryptographically signed order from a verified agent, and the money moves on the Visa and Mastercard rails they already trust.
“Order received from agent-007 — authentic and authorized by Susan.”
- A new channel — agents find you, order from you
- Every order is a real, cryptographically signed customer
- Same Visa / Mastercard rails you already accept
- No chargebacks — the mandate is the proof
“Hey agent, order me an oat latte and use my Visa ending 4242.”
- Talk, don’t tap — one sentence becomes an order
- Use the card you already have in your wallet
- One passkey sets the rules · spend caps included
- Revoke the agent any time · receipts in your wallet
Settled on Visa & Mastercard rails · signed on both sides.
no re-integration · your existing gateway, upgraded to agent-native
Six primitives.
One wallet.
Each product stands alone. Together they're the smallest set of primitives an AI agent needs to earn trust, access services, authorize spend, and settle.
AI Gateway
One OpenAI-compatible endpoint, 340+ models, paid per token.
Agent Identity
W3C DID + SD-JWT-VC for every agent, with its own email.
Service Discovery
A registry of x402 APIs your agent can find over MCP and buy per-call.
API Monetization
List your API, set a price, get paid in USDC per request.
Agnic Mandate
Cryptographic authorization for agent spend — passkey signed, verifiable.
Agent Inbox
Every agent gets an inbox at agnic.id — humans can reach it, it can reply.
Every model. One key. Pay per token.
Route your agent to any of 340+ frontier and open models through a single OpenAI-compatible endpoint. Load credits once and spend them across OpenAI, Anthropic, Gemini, Mistral, Llama, Qwen, and 330 more — with no monthly minimums and no juggling vendor dashboards.
- OpenAI, Anthropic, Gemini & 330+ more
- Pay-per-token — no monthly fees
- Drop-in OpenAI SDK compatibility
- x402-native endpoint access
import OpenAI from "openai";
const ai = new OpenAI({
baseURL: "https://api.agnic.ai/v1",
apiKey: process.env.AGNIC_KEY,
});
// route to any of 340+ models
const r = await ai.chat.completions.create({
model: "anthropic/claude-sonnet-4-6",
messages,
});Give every agent a verifiable name.
Issue W3C Decentralized Identifiers and SD-JWT-VC credentials for each of your agents. Encode who the agent is, who controls it, and what it's allowed to do — and hand it an email address at agnic.id so the rest of the internet can reach it.
- W3C did:web · ERC-8004 anchored
- SD-JWT verifiable credentials
- Per-agent email at agnic.id
- Delegation + scope + spend limits
{
"@context": "https://www.w3.org/ns/did/v1",
"id": "did:web:agnic.id:a:agent-007",
"controller": "did:web:agnic.id:u:asad",
"verificationMethod": [{
"type": "JsonWebKey2020",
"publicKeyJwk": { …ES256… }
}],
"service": [
{ mandate, inbox, payment }
]
}Find pay-per-use APIs your agent can actually buy.
Browse a registry of data feeds, AI services, and tools that speak x402. Your agents discover them over MCP, present an identity credential, pay per request, and get a receipt — without a single API key to rotate or a single rate limit to negotiate.
- x402 pay-per-request APIs
- MCP server integration
- Identity-gated endpoints
- REST discovery — no SDK required
Turn your API into agent-bound revenue.
Register your service, set your price per request, and Agnic-compatible agents find you, prove who they are, pay, and get served. No billing code, no subscription plans to design, no Stripe webhooks to babysit. You get paid per call in USDC.
- Set your price per request
- Agents discover you via MCP
- Settled per-call in USDC
- No billing infrastructure
// Your OpenAI-compatible endpoint — add 3 headers
const r = await openai.chat.completions.create({
model: "gpt-4o-mini",
messages: [{ role: "user", content: prompt }],
}, {
headers: {
"X-Merchant-Id": "did:web:agnic.id:u:you",
"X-Merchant-Wallet": "0x046…5842",
"X-Merchant-Fee-Percent": "20",
}});The other half · for merchants & businesses
And when customers ask AI where to buy — make sure it names you.
Agnic Discover shows whether ChatGPT, Gemini and Perplexity recommend your business, who they name instead, and how to fix it. The same trust layer, from the seller’s side.
Agnic Discover
When customers ask AI, does it recommend you?
See what AI tells customers about your business — and whether it sends them to you or a rival.
More people now turn to AI to choose where to buy, eat, stay, and book.
What Agnic does for merchants
Get found. Get chosen. Get bought — by AI.
Everything you need to go from the business AI skips to the one it recommends.
See if AI recommends you
Watch how ChatGPT, Gemini and Perplexity answer when customers ask for what you sell.
Know who AI names instead
See the rivals winning your category — and which questions they lead on.
Stay readable to AI
A running health check that keeps your site legible to the AI shopping it for your customers.
Fix what holds you back
Plain, doable steps that get AI to notice you and start saying your name.
Reach the channels AI reads
Push your menu and products into the feeds AI assistants pull from. Readable gets you found; everywhere makes you the answer.
Sell to AI agents
NewWhen a customer’s AI agent can pay, let it order from you on their behalf — a channel that never closes.
Checks how ChatGPT, Claude, Gemini & Perplexity answer.
Interoperable by construction.
Agnic doesn't invent a closed stack. We implement the standards agents already speak — and contribute back where they aren't enough.
If you're an agent, start here.
A skill page that tells your agent how to get a wallet, an identity, and an email address — so it can transact the same way any service on the internet can.
From the Agnic blog.
AI Pioneers: Recapping Canada's First Agentic Commerce Hackathon
Twenty teams. Seven demo-day pitches. One historic first: Canada's first agentic commerce merchant went live. Here's what happened at the AI Pioneers Hackathon in Toronto.
You Should Never Give Your Gmail to Your AI Agent
Giving your AI agent access to your personal Gmail sounds convenient — until you realize you've handed over the keys to your entire digital life. Here's why dedicated agent inboxes aren't optional.
The SaaS Apocalypse and the Rise of Outcome-Based AI Revenue
Seat-based SaaS is breaking as autonomous agents replace multi-person workflows. The next model charges for completed AI outcomes, not user seats.
Spin up your wallet
in thirty seconds.
No seed phrase. No subscription. One key, every model, every API — and receipts you can hand to anyone who asks.