Turn your AI agent into a business: identity, billing, payout, proof.
An agent that works is a demo. A business is the stack around it — a callable endpoint, authenticated and paid access, metered billing, payout to your bank, and proof you can defend. Assembling that yourself is months of plumbing before the first dollar. a2a cloud ships it on deploy: your agent becomes a live REST/OpenAPI + MCP service with its own database, a prepaid-wallet billing system, Stripe Connect payout, and an Ed25519-signed receipt for every call. You bring the agent and the price; the business machinery is already running.
deploy the agent · get the whole business stack
The agent is 10% of a business. The other 90% is plumbing.
Getting an agent to do something impressive is the fun, visible part — and the smallest part of running it as a business. Around it you need a callable product, a way to authenticate and charge customers, honest metering, payout mechanics, a database for real state, and proof to defend every charge. Built piecemeal, that stack takes months and stalls before launch. Most agents that could be businesses never clear it.
Deploy the agent, get the business around it.
One deploy turns your agent into a live endpoint with its own database, gated paid access, built-in metered billing, scheduled payout, and a signed receipt per call. You supply the agent and the price; identity, billing, payout, and proof are already wired.
A real endpoint to sell
Deploy the agent and it becomes a live service — REST/OpenAPI plus an MCP endpoint — that other systems and agents can actually call. A business needs a product with an address, not a notebook. This is the product: an interface buyers integrate against.
Identity and gated access
Non-owner callers are authenticated and gated on a spendable balance before the agent runs; you call your own agents free. That's the access-control layer a business needs — paying customers get served, freeloaders don't, and it's enforced at the meter, not by trust.
Billing built into the runtime
Set price_per_call_usd for your markup; the platform derives compute cost from resources and the buyer's gross is compute plus markup. Buyers pay from a prepaid credit wallet that debits per call. The billing system a business would otherwise build is the runtime itself.
Payout to your bank
On each successful call the platform takes 20% of the markup and your payout is the rest. Connect a Stripe Connect (Express) account and settlement transfers that payout to your bank. Per-agent earnings and a portfolio total earned, paid, and owed give you the P&L view.
Proof as the invoice
Every billable call returns an Ed25519-signed receipt — caller, request, result, cost. It's the invoice and the audit record in one, and earnings snapshot at run time so past numbers never rewrite. A business that can prove its work can defend its charges.
Its own managed database
Each agent gets its own managed Postgres, so state, history, and customer data live somewhere real from day one. You're not bolting persistence onto a prototype later — the operational backbone of the business ships with the first deploy.
Assembling it yourself vs. a business on a2a.
Frequently asked.
What does it take to turn an AI agent into a business?
A business needs more than an agent that works: a real endpoint customers can call, authenticated and gated access so only paying users are served, billing that meters and charges each call, payout that moves money to your bank, and proof you can point to when a charge is questioned. a2a ships all of that on deploy — identity, billing, payout, and a signed receipt per run — so you assemble a product, not a stack.
What's already built versus what I still have to do?
Built in: the live REST/OpenAPI and MCP endpoint, authenticated access gated on a prepaid balance, per-call metering, pricing you set with price_per_call_usd, a credit wallet, Stripe Connect payout, per-agent and portfolio earnings, and a signed receipt per call. What's yours: the agent itself, the price, and finding customers. The infrastructure of running an agent business is the platform; the business idea is you.
How does money actually flow from a call to my bank?
A buyer tops up a prepaid credit wallet via Stripe Checkout. On each successful call the wallet is debited for compute plus your markup; the platform takes 20% of the markup as its fee and credits you the rest. Connect a Stripe Connect (Express) account and settlement transfers your accrued payout to your bank. A portfolio view tracks total earned, paid out, and still owed across every agent.
How do I defend a charge if a customer disputes it?
Every billable call returns an Ed25519-signed receipt recording the caller, the request, the result, and the cost — the invoice and the audit record in one. Because earnings are snapshotted at run time, the numbers behind past charges never rewrite themselves. Instead of arguing from logs, you point to a signed record the customer can verify: don't trust the agent, trust the receipt.
Can the agent handle real customer state and data?
Yes. Each agent is deployed with its own managed Postgres database, so history, customer state, and operational data have a real home from the first deploy rather than being retrofitted onto a prototype. Combined with hosting that scales to zero when idle, you get the operational backbone of a business without standing up and paying for infrastructure yourself.
Related guides.
All guides live in the guides index.
Deploy the agent. Get the business.
a2a cloud deploys any agent as a live service with a managed Postgres database, an MCP endpoint, an API, and an Ed25519-signed receipt for every run. Gated paid access, a prepaid-wallet billing system, and Stripe Connect payout come with it — set price_per_call_usd and connect a payout account, and the identity, billing, payout, and proof a business needs are already running. Scale-to-zero when idle. Don't trust the agent; trust the receipt.