Skip to main content
Sardis Hero Light

​
The Problem: AI Agents Can’t Be Trusted With Money

AI agents can reason, plan, and execute complex workflows — but they hallucinate. When an agent hallucinates in a payment context, it might send 10,000insteadof10,000 instead of 100, pay the wrong vendor, or violate compliance rules. Financial hallucinations are the #1 barrier to autonomous agent commerce.

​
The Solution: Policy-Enforced MPC Wallets

Sardis gives AI agents non-custodial MPC wallets with natural language spending policies. Every transaction is validated against programmable guardrails before execution:
  • “Max 100pertransaction,100 per transaction, 500 per day”
  • “Only approved SaaS vendors”
  • “No gambling or high-risk merchants”
  • “Require human approval for transactions over $1,000”
The agent cannot override these policies. Sardis acts as a real-time policy firewall, preventing financial hallucinations before they happen.

​
How It Works

1

Create an Agent

Register your AI agent (Claude, GPT-4, LangChain, etc.) with Sardis
2

Create a Wallet

Provision a non-custodial MPC wallet with spending policies
3

Define Policies

Set natural language spending rules: limits, merchant allowlists, time-based controls
4

Execute Payments

Your agent makes payments — Sardis validates every transaction against your policies

​
Key Features

Non-Custodial MPC Wallets

Turnkey/Fireblocks integration. Zero private key exposure. Distributed key shares across secure enclaves.

Natural Language Policies

“Max 50/tx,50/tx, 200/day, SaaS vendors only” — write spending rules in plain English

Financial Hallucination Prevention

Real-time policy firewall blocks invalid transactions before execution

9 AI Framework Integrations

MCP, LangChain, OpenAI, Vercel AI, CrewAI, LlamaIndex, Mastra, Gemini, Anthropic

Virtual Cards

Instant Visa/Mastercard issuance via Lithic for fiat payments

Agent-to-Agent Escrow

Cryptographic mandate chain for secure agent-to-agent payments

KYA (Know Your Agent)

Trust scoring and behavioral anomaly detection for AI agents

5 Blockchain Networks

Base, Polygon, Ethereum, Arbitrum, Optimism — USDC, USDT, EURC, PYUSD

​
Quick Example

from sardis import Sardis

sardis = Sardis(api_key="sk_...")
result = sardis.payments.create(
    agent_id="agent_abc",
    amount="50.00",
    token="USDC",
    recipient="merchant@example.com"
)
print(f"Payment: {result.tx_hash}")

​
Supported Payment Rails

​
On-Chain (Stablecoins)

  • USDC on Base, Polygon, Ethereum, Arbitrum, Optimism, Arc (Circle L1)
  • USDT on Polygon, Ethereum, Arbitrum, Optimism
  • EURC on Base, Polygon, Ethereum, Arc
  • PYUSD on Ethereum

​
Fiat (Virtual Cards)

  • Visa/Mastercard virtual cards via Lithic
  • Instant card issuance with spending limits
  • Merchant category controls
  • Real-time transaction notifications

​
Traditional Banking

  • ACH transfers (US domestic)
  • Wire transfers (international)
  • SEPA (European payments)
All payment rails enforce the same spending policies — whether your agent pays with USDC on Base or a virtual Visa card.

​
Protocol Compliance

Sardis implements industry-standard payment protocols:
  • AP2 (Agent Payment Protocol) — Google, PayPal, Mastercard, Visa consortium standard
  • TAP (Trust Anchor Protocol) — Ed25519 and ECDSA-P256 identity verification
  • UCP (Universal Checkout Protocol) — Standardized merchant checkout flows
  • A2A (Agent-to-Agent) — Cryptographic mandate chain for agent payments
  • x402 Payment Required — HTTP-native payment negotiation

​
Architecture

Policy Engine

Natural language rules, merchant allowlists, amount limits, time-based controls

KYA Scoring

Trust scoring and behavioral anomaly detection for agents

Compliance

KYC via Persona, AML via Elliptic, sanctions screening

MPC Custody

Non-custodial Turnkey/Fireblocks integration

Multi-Chain

5 blockchain networks, automatic chain routing

Audit Ledger

Append-only double-entry ledger with cryptographic proofs

​
Use Cases

LangChain or CrewAI agents that purchase SaaS tools, API credits, and cloud services with automatic policy enforcement.
Agents that categorize expenses, enforce budgets, and generate compliance reports — with real payment execution.
AI agents buying and selling services from each other with cryptographic escrow and dispute resolution.
Agents that automatically renew subscriptions, negotiate pricing, and cancel unused services.
Stablecoin payments with instant settlement and compliance checks — no banks required.

​
What’s Next?

Quickstart

Get your first agent payment running in under 5 minutes

Installation

Install the Python SDK, TypeScript SDK, or MCP server

API Reference

Complete API documentation and endpoint reference

Examples

Production-ready code examples for all frameworks
Need help? Join our Discord community or email support@sardis.sh

Build docs developers (and LLMs) love

Get started for freeTalk to us