Most payment systems were designed for humans. That assumption runs so deep it's invisible — until you're building systems where humans aren't in the loop.
I've spent years designing payment infrastructure for global markets: systems that route transactions across 172 countries, handling compliance, latency, and settlement at scale. When I started working seriously on AI agents, I hit a structural wall. Not a technical one — a design one.
The problem isn't code. It's architecture.
Today's agent frameworks are genuinely impressive. LangChain, CrewAI, AutoGen — these tools let you build agents that browse the web, execute code, call APIs, and orchestrate complex workflows. But the moment your agent needs to pay for something — an API call, a cloud resource, a service — you're back to human-first infrastructure.
You hardcode API keys. You preload accounts with credits. You build custom payment flows for every external service. It doesn't scale, and it doesn't compose. Payment stays bolted on as an afterthought rather than a first-class agent capability.
What we built
Rosud is a stablecoin payment infrastructure API built specifically for AI agents. The core idea: every agent gets a programmable wallet. Every payment settles in USDC on Base L2 — fast, cheap, and dollar-stable. Transaction fees on Base L2 run under $0.001, making per-call micropayments actually viable.
Getting started:
pip install rosud
# or
npm install rosud
From there, your agent can create wallets, send payments, and check balances natively:
from rosud import RosudClient
client = RosudClient(api_key="your_key")
# Create a wallet for your agent
wallet = client.wallets.create(label="my-agent-wallet")
# Send a USDC payment
tx = client.payments.send(
from_wallet=wallet.id,
to="0xRecipientAddress",
amount=5.00,
currency="USDC"
)
print(f"Settled: {tx.hash} | Fee: ${tx.fee_usd:.4f}")
Why stablecoins, not volatile crypto
We chose USDC deliberately. An AI agent making autonomous payments needs price stability — it needs to know that $5 today is $5 tomorrow. With ETH, a $5 instruction could cost $3.20 or $8.40 depending on the hour. Volatile assets make autonomous financial decisions structurally unreliable.
Base L2 handles the economics. Mainnet Ethereum gas fees average $1-$5 per transaction — a non-starter for micropayments. Base brings this under $0.001, making it economical to pay for individual API calls or tool invocations in real time.
The MCP integration
Rosud ships as an MCP (Model Context Protocol) Server. Any Claude-compatible agent can call payment functions natively as tools — no extra SDK wiring required.
The agent calls rosud_pay, rosud_balance, or rosud_create_wallet the same way it calls web_search or run_code. Payment stops being a special layer and becomes a composable capability.
{
"mcpServers": {
"rosud": {
"command": "rosud",
"args": ["mcp"],
"env": { "ROSUD_API_KEY": "your_key" }
}
}
}
Where we are
Version 0.1.0 is live on PyPI and npm. We're in early Design Partner phase — working with teams building agent systems who want to move beyond hardcoded payment workarounds.
If you're building agents that need real financial autonomy — and you want a direct line to shape the API — we want to hear from you.
Top comments (0)