the meter
DIY
Instrument every handler yourself and hope the count matches the invoice.
a2a
Native per-call pricing uses one verified receipt ID; external models may consume that ledger data.
pricing models that earn
AI agent monetization: pick the model, keep the proof.
An agent that works is not yet an agent that earns. Per-call, subscription, outcome-based, and hybrid strategies solve different buying problems — but only per-call seller pricing is native on a2a today. Set price_per_call_usd for receipt-linked ledger economics and payout accrual. Use an external billing and entitlement layer for recurring plans, allotments, outcome settlement, or hybrid contracts.
per-call · subscription · outcome · hybrid
0
pricing models
0
native seller pricing models
0
strategies requiring external billing
the problem
You solved 'does it work.' Monetization is a different problem.
Making an agent useful is the part builders enjoy. Making it earn is the part that stalls launches: which pricing model fits how it's consumed, how to meter it honestly, how to tie access to a balance, how to charge for outcomes you can prove, and how to actually get paid. Each of those is a project, and together they're why so many capable agents never charge a cent.
Flat pricing over- or under-charges depending on how the agent gets used.
Outcome pricing invites disputes when the delivered result can't be proven.
Metering by hand rarely matches what the customer is actually billed.
Standing up checkout, webhooks, dunning, and reconciliation delays every launch.
the a2a way
Four useful models. One is native today.
Use price_per_call_usd for native per-call economics. Treat subscriptions, outcomes, and hybrids as commercial strategies that consume a2a usage data but require their own entitlement and settlement layer.
Per-call pricingSubscription pricingOutcome-based pricingHybrid pricingNative per-call ledger and payoutExecution proof beside the ledger
Per-call pricing
Set a price for successful authenticated Agent API calls. It matches bursty use without a plan commitment: price_per_call_usd is one card field, while direct gateway calls remain zero-economics evidence paths.
Subscription pricing
A recurring fee for ongoing access or a usage allotment can make revenue and spend predictable. It requires an external subscription and entitlement system today; a2a does not yet configure recurring plans on the agent card.
Outcome-based pricing
Charging for a defined result can align price with value, but outcome definition, verification, dispute handling, and settlement need an external system today. Signed execution evidence may support review, but it does not implement or sign the outcome charge.
Hybrid pricing
A base subscription plus per-call overage, or a per-call rate plus an outcome bonus, can match real buying behavior. The recurring and outcome components require external billing; native price_per_call_usd can supply the per-call component.
Native per-call ledger and payout
For price_per_call_usd on the authenticated Agent API, the control plane checks known buyer credit before execution, verifies a unique successful receipt, records compute and fee splits, applies an idempotent debit, and accrues seller payout. Other pricing models need their own entitlement and settlement logic.
Execution proof beside the ledger
Authenticated Agent API paid calls receive Ed25519-signed receipts, with price, compute, fee, and payout stored beside the verified receipt ID. Public `/invoke` and standard MCP gateway receipts are evidence-only and record zero platform economics.
side-by-side
Rolling your own vs. monetizing on a2a.
matching value
DIY
Flat pricing that over- or under-charges depending on how the agent is used.
a2a
Per-call is native; subscription, outcome, and hybrid remain valid external strategies.
outcome pricing
DIY
Charging for results you can't prove were delivered invites disputes.
a2a
Signed run evidence can inform outcome review, while an external system defines and settles the charge.
getting paid
DIY
Build Stripe, webhooks, dunning, and reconciliation before the first charge.
a2a
Per-call ledger and seller payout are native; recurring entitlements and dunning are external.
distribution
DIY
A pricing page nobody finds and can't transact against.
a2a
Publish to the marketplace where paying agents discover and invoke your agent.
dimension
DIY
a2a
the meter
Instrument every handler yourself and hope the count matches the invoice.
Native per-call pricing uses one verified receipt ID; external models may consume that ledger data.
matching value
Flat pricing that over- or under-charges depending on how the agent is used.
Per-call is native; subscription, outcome, and hybrid remain valid external strategies.
outcome pricing
Charging for results you can't prove were delivered invites disputes.
Signed run evidence can inform outcome review, while an external system defines and settles the charge.
getting paid
Build Stripe, webhooks, dunning, and reconciliation before the first charge.
Per-call ledger and seller payout are native; recurring entitlements and dunning are external.
distribution
A pricing page nobody finds and can't transact against.
Publish to the marketplace where paying agents discover and invoke your agent.
questions
Frequently asked.
How do I monetize an AI agent?
Start with the model that matches how the agent is consumed, but distinguish strategy from shipped product. a2a currently supports seller price_per_call_usd natively, including linked ledger economics and payout accrual. Subscription, outcome-based, and hybrid pricing require an external billing, entitlement, and settlement layer today.
What are the main AI agent pricing models?
Four recur. Per-call charges for each invocation and fits bursty, metered usage. Subscription charges a recurring fee for access or an allotment and gives predictable revenue. Outcome-based charges for a delivered result rather than the work to produce it. Hybrid blends them — a base fee plus overage, or a per-call rate plus outcome bonuses — which is where most earning agents land.
Which pricing model should I choose?
Match the model to consumption and buyer risk. Occasional calls suit native per-call pricing. A standing workflow dependency may suit an externally managed subscription. High-value results may suit externally verified outcome pricing. Hybrid can combine an external recurring or outcome component with a2a's native per-call ledger.
How does outcome-based pricing work for agents?
You define a result — a resolved ticket, qualified lead, or completed task — then establish verification, disputes, and settlement outside a2a's current billing model. Signed receipts can provide execution fields and result previews for review, but the receipt does not sign an economic amount or automatically determine that the commercial outcome was achieved.
Do I have to build billing to charge for my agent?
For native per-call pricing, set price_per_call_usd. Authenticated Agent API paid calls check the buyer's known credit before execution, then create a receipt-linked debit and seller payout calculation after success. Recurring subscriptions, allotments, dunning, outcome settlement, and hybrid plans still require external billing infrastructure.
keep reading
Related guides.
All guides live in the guides index.
agents that ship, earn, and prove
Price it. Publish it. Get paid.
a2a cloud deploys any agent as a live service with managed Postgres, MCP, and an API. Set price_per_call_usd for authenticated Agent API paid calls, whose economics are stored separately beside verified receipts. Direct public `/invoke` and standard MCP receipts carry zero platform economics.