A new revenue line for your SaaS.
Your customers' agents are already using your software. Right now you can't see them, control them, or charge them. Agent Checkout turns agent usage into metered, billable revenue. Drop-in SDK. Your accounts. Your pricing.
Agents are a buyer channel your billing system cannot see.
Your product already has workflows customers want agents to run: research, enrichment, analysis, export, routing, remediation, and review. Seat pricing was built for people logging in. Agents buy work.
Agent Checkout lets you sell that work directly. You define the billable actions, attach credits or token prices, let customers approve which agents can act, and charge only when approved work succeeds.
The result is a new merchant revenue line on top of your existing seats, plans, and account contracts.
The work is happening. The invoice is not.
A customer connects an agent to your software. It runs the workflow your sales team already positions as premium work. Under a seat model, the usage looks like one busy account and the customer pays the same monthly plan.
With Agent Checkout, that same workflow becomes a priced action. The customer preloads credits or buys a workflow pack, the agent spends against an approved cap, and every successful run creates a billing record.
You keep the customer relationship. You keep your pricing logic. Agent Checkout adds the missing usage ledger between agent work and merchant revenue.
Seats measure people. Agent Checkout meters work.
To sell agent usage, you need more than an API key and an invoice line. You need customer approval, scoped agent identity, priced actions, success-based metering, and evidence finance can reconcile.
That is the part Agent Checkout packages for the merchant. Your team exposes the workflows. Nitrograph handles the agent commerce controls around them.
You can pilot with a few high-value actions, then expand into additional workflow packs as customers prove demand.
Agent Checkout turns agent work into revenue.
Customer-approved agents
Customers approve named agents, set spend caps, choose allowed actions, and revoke access anytime.
Successful work only
Agent Checkout meters approved billable actions only when the work succeeds. Failed calls do not charge.
Evidence finance can use
Each closed cart includes the customer, grant, action, token cost, result, and invoice reference.
Cart opened · agent: claude-acme-ops · customer: northstar · cap: 500 tokens + action · contract_analysis · 25 tokens · success + action · contract_analysis · 25 tokens · success + action · clause_extraction · 12 tokens · success + action · contract_analysis · 25 tokens · FAILED -> not metered Cart closed · 62 tokens billed · billing evidence generated
The part you cannot build alone is cross-merchant reputation.
Any engineering team can eventually build basic agent metering for their own product. That is useful plumbing, but it only sees your own traffic.
When an agent reaches your software, you need to know if it is acting for a real customer, pays reliably, respects limits, and avoids disputed usage. A single SaaS merchant only sees one slice of that behavior.
Nitrograph sits across every merchant running Agent Checkout. We see agent behavior across transactions, caps, denials, retries, settlements, and disputes. That lets merchants trust approved agents faster and bill with less ambiguity.
The more merchants meter through Agent Checkout, the better the reputation data gets for every merchant in the network.
Turn agent traffic into revenue.
Agent Checkout is a drop-in SDK. You keep your existing accounts, your existing pricing, and your existing customers. You add a new revenue line from a user class you're currently serving for free.
We take a cut of metered agent transactions. You keep the rest. No subscription, no seat license, no upfront platform fee. You only pay when agents generate revenue.
Once Checkout is live, agents can find approved work themselves.
Agent Checkout turns your workflows into approved, priced, agent-callable actions. Once those actions are live in your stack, agents can find the merchant-approved work they are allowed to run and pay through the same controls, caps, and evidence trail.
Explore the merchant path →