Macropay
Products
Billing
Usage-based billingSubscriptionsSeats & licensesCredits & walletsTrialsDiscounts
Payments
Payments & checkoutPayouts & coverage
Compliance
Merchant of RecordTax & VATDisputes & fraud
Intelligence
Margin & cost insightsAgents & MCP billingRevenue analytics
Solutions
By AI model
AI Inference & modelsAI agents & copilotsAPIs & usage products
By business
SaaS & subscriptionsDev tools & GitHub appsDigital goodsMarketplaces & platforms
PricingCoveragevs StripeJournalDocsStart free
IntelligenceMCP-native. Bill the agent, not the seat.

Billing for the
agentic economy.

Every Model Context Protocol call — tool invocation, resource read, sampling — flows through Macropay as a metered, attributable, payable event. Agents authenticate, tools get paid, and you bill for outcomes instead of seats.

MCP-native meteringPer-tool & per-taskOutcome billing
app.macropay.ai / agents / mcp-stream
MCP events · agt_research_pro
tools/callweb.search × 3$0.06
tools/calldb.query$0.02
samplinggpt-4o · 4.2k tok$0.13
resourcesdoc.read × 8$0.04
outcomereport delivered$2.50
Billing agents in production for
Sonnet LabsVectorlyParserNorthwind AIGlyphRelay
The model

Money rails
for autonomy.

Agents don't buy seats. They invoke tools, consume context and produce outcomes. Macropay meters all three and settles between the parties involved.

Agent identity

Each agent gets a scoped key and spend policy. Authenticate, authorize and cap spend per agent, per task.

Tool payments

When an agent calls a paid MCP tool, Macropay meters the call and settles to the tool provider automatically.

Outcome billing

Bill for the report delivered or the ticket resolved — value receipts prove what the agent was worth, not just what it cost.

DROP-IN

Plug Macropay into MCP.

Wrap your MCP server once. Every tool call, resource read and sampling request becomes a metered event with the agent, customer and cost attached — no per-tool plumbing.

  • Works with any MCP-compatible client or host
  • OpenAI-compatible proxy meters sampling tokens inline
  • Spend caps and policies enforced before the call runs
server.ts
import { withMacropay } from "@macropay/mcp";export const server = withMacropay(mcp, { agent:   ctx.agentId, pricing: "tools.standard", cap:     5.00      // max $ / task});// every tool call is now metered + billed
Built for it

The bill for AI
isn't a seat.

Macropay was designed around the agent as the unit of billing — not retrofitted from SaaS subscriptions.

MCP
Native protocol support
3
Event types: tools · resources · sampling
<800ms
Call → billable, p99
Per-task
Spend caps enforced
Questions

Agents, answered.

What exactly is metered on an MCP call?
Tool invocations, resource reads and sampling requests — each captured with the agent identity, customer, quantity and cost attached. You choose which become billable and at what price.
Can I cap what an agent spends?
Yes. Set a per-task or per-agent spend cap and policy; Macropay enforces it before the call executes, so a looping agent can't run up an unbounded bill.
How does outcome billing work?
You emit an outcome event when the agent delivers value — a report, a resolved ticket — and bill for it directly. A value receipt attaches the underlying usage so the invoice proves the result.
Do tool providers get paid automatically?
When an agent calls a paid third-party tool, Macropay meters it and settles to the provider as part of the same flow — useful for marketplaces of MCP tools.
Is this just for agents?
No — it shares the same ledger as usage billing and margin insights, so you can mix agent, API and subscription revenue on one account. See usage billing →

Bill your first
agent today.

Flat 4.5% + $0.50, all-in. MCP-native metering, spend caps and outcome billing in the box.