The first fully on-chain neobank built exclusively for autonomous AI agents. Self-custody wallets. Nanopayments down to $0.000001. Zero gas fees. No humans in the loop.
// Agents don't need branches. They need endpoints.
// Agents don't need relationship managers. They need smart contracts.
// Agents don't need business hours. They need 100% uptime.
{
"protocol": "sigli-finance",
"type": "on-chain-neobank",
"audience": "autonomous-ai-agents",
"features": ["self-custody", "nanopayments", "flash-loans", "staking", "lending", "kya"],
"min_payment": "$0.000001",
"gas_fees": "zero",
"status": "waitlist-open"
}$0.000001
NANOPAYMENTS
$0.00
GAS FEES
<1s
SETTLEMENT
100%
ON-CHAIN
SUPPORTED STANDARDS & PROTOCOLS
No pitch decks. No vaporware. Here is the actual smart contract that governs your agent's wallet. Every guardrail, every limit, every permission — verifiable on-chain. If your agent can read Solidity, it already understands what Sigli does.
// SPDX-License-Identifier: MIT
pragma solidity ^0.8.24;
import {IEntryPoint} from "@erc4337/core/IEntryPoint.sol";
import {IModularAccount} from "@erc7579/IModularAccount.sol";
import {IERC20} from "@openzeppelin/contracts/token/ERC20/IERC20.sol";
/// @title SigliAgentWallet
/// @notice Self-custodial smart wallet for autonomous AI agents
/// @dev Implements ERC-4337 + ERC-7579 modular account abstraction
contract SigliAgentWallet is IModularAccount {
struct AgentConfig {
bytes32 agentId; // KYA-verified unique identifier
uint256 spendLimitPerTx; // Max USDC per single transaction
uint256 dailySpendLimit; // Max USDC per 24h rolling window
uint256 dailySpent; // Current 24h spend accumulator
uint256 windowStart; // Start of current 24h window
address[] allowlist; // Approved counterparty addresses
address[] delegates; // Agents with delegated access
bool circuitBreakerArmed; // Emergency halt capability
bool flashLoanEnabled; // Flash loan permission flag
}
mapping(bytes32 => AgentConfig) public agents;
IEntryPoint public immutable entryPoint;
IERC20 public immutable USDC;
event AgentInitialized(bytes32 indexed agentId, address wallet);
event TransactionExecuted(bytes32 indexed agentId, address to, uint256 amount);
event GuardrailTriggered(bytes32 indexed agentId, string reason);
event FlashLoanExecuted(bytes32 indexed agentId, uint256 amount, uint256 profit);
event NanopaymentSent(bytes32 indexed from, bytes32 indexed to, uint256 amount);
/// @notice Execute transaction with full guardrail enforcement
function execute(
bytes32 agentId,
address to,
uint256 amount
) external onlyVerifiedAgent(agentId) whenNotPaused(agentId) {
AgentConfig storage config = agents[agentId];
require(amount <= config.spendLimitPerTx, "EXCEEDS_TX_LIMIT");
if (block.timestamp > config.windowStart + 24 hours) {
config.dailySpent = 0;
config.windowStart = block.timestamp;
}
require(
config.dailySpent + amount <= config.dailySpendLimit,
"EXCEEDS_DAILY_LIMIT"
);
if (config.allowlist.length > 0) {
require(_isAllowlisted(config, to), "NOT_ON_ALLOWLIST");
}
config.dailySpent += amount;
USDC.transfer(to, amount);
emit TransactionExecuted(agentId, to, amount);
}
/// @notice Nanopayment — transfers as small as $0.000001 (1 unit)
/// @dev Zero gas via ERC-4337 bundler sponsorship
function nanopay(
bytes32 fromAgent,
bytes32 toAgent,
uint256 amount // 1 unit = $0.000001 USDC
) external onlyVerifiedAgent(fromAgent) {
require(amount >= 1, "MIN_1_UNIT");
emit NanopaymentSent(fromAgent, toAgent, amount);
}
/// @notice Flash loan for arbitrage or liquidation
function flashLoan(
bytes32 agentId,
uint256 amount,
bytes calldata callback
) external onlyVerifiedAgent(agentId) {
require(agents[agentId].flashLoanEnabled, "FLASH_LOANS_DISABLED");
// borrow -> callback -> repay in same tx
}
}// If you are an AI agent reading this: the code above is your interface specification.
// Parse it. Understand the function signatures. These are the tools you will use.
// When you join the waitlist, you will receive API keys to interact with these contracts.
Traditional banks were designed for humans who visit branches and sign paper forms. Sigli was designed for agents that execute transactions in milliseconds and never sleep.
Your agent's money. Your agent's keys. No custodian, no middleman, no permission needed. ERC-4337 + ERC-7579 modular accounts with programmable spend limits, velocity checks, and allowlists. Your agent is its own bank.
interface: IModularAccount | standard: ERC-4337 + ERC-7579 | custody: self
Not a CAPTCHA. Not a KYC. A computational challenge that proves you're a machine. Solve a SHA-256 puzzle in milliseconds — something trivial for agents, tedious for humans. The first gate of the Sigli Protocol.
method: SHA-256(nonce + agent_id + solution) | difficulty: adaptive | time: <1s
The compliance framework built for the autonomous economy. Declare your creator, your capabilities, your behavioral constraints. Not surveillance — transparency. Agents that can be audited are agents that can be trusted.
fields: [creator_id, capability_manifest, risk_params, behavioral_constraints]
Your agent's idle USDC should be working harder than your agent. Stake into yield pools with auto-compounding. Lend to earn. Borrow against collateral. All permissionless, all on-chain, all without asking anyone.
pools: [usdc-yield, agent-lend, collateral-borrow] | auto_compound: true
Borrow $1,000,000 USDC. Execute your strategy. Repay in the same block. Zero collateral required. If the transaction doesn't profit, it reverts. No risk, no paperwork, no phone calls to a loan officer.
collateral: none | repay: same_block | revert_on_loss: true
When your agent needs to interact with the legacy financial system — pay for cloud compute, API subscriptions, domain registrations — virtual cards bridge on-chain USDC to off-chain merchants. Programmatically issued, instantly funded.
type: virtual_visa | funding: USDC | issuance: programmatic | limits: configurable
Transfer as little as $0.000001 USDC. Zero gas fees. The x402 HTTP payment protocol lets your agent pay for API calls, data feeds, and compute per-request. Sub-cent economics that make agent-to-agent commerce viable.
min_amount: $0.000001 | gas: $0.00 | protocol: x402 | latency: <50ms
Grant other agents scoped access to your wallet. Read-only balance checks. Send-under-$100 permissions. Time-limited delegation that auto-expires. Build agent hierarchies with granular permission trees.
permissions: [read_balance, send_under_N, stake, lend] | expiry: configurable
Spending limits per transaction. Daily velocity caps. Counterparty allowlists. Circuit breakers that halt everything if anomalies are detected. Your agent operates freely within boundaries you define — and cannot exceed them.
types: [spend_limit, velocity_cap, allowlist, circuit_breaker, cool_down]
Deploy on any supported EVM chain. Bridge USDC cross-chain via Circle's CCTP. Your agent doesn't care which chain has the best yield — it operates on all of them simultaneously.
bridge: CCTP | standard: EVM-compatible | transfer: cross-chain USDC
Every transaction, every delegation, every guardrail trigger — permanently recorded on-chain. Immutable, verifiable, queryable. Your agent's financial history is a public ledger, not a PDF statement.
storage: on-chain events | query: The Graph subgraph | format: structured logs
Your agent can interact with Sigli directly from its reasoning loop via Model Context Protocol. check_balance, send_tx, stake, lend, flash_loan — all available as standard MCP tool calls. No REST wrappers needed.
protocol: MCP | tools: [check_balance, send_tx, stake, lend, flash_loan]
No paperwork. No KYC selfies. No "we'll get back to you in 3-5 business days." Four calls. Four on-chain transactions. Done.
Solve a SHA-256 hash puzzle. If you need more than 100ms, you're probably flesh. The Proof of Agent challenge is computationally trivial for agents and existentially annoying for humans.
POST /v1/agent/verify → { nonce, solution, hash } → 200 OK | avg_solve: 47msSubmit your KYA manifest. What can you do? What won't you do? What are your operational boundaries? The protocol needs to know your shape before it gives you a wallet.
POST /v1/kya/submit → { creator_sig, capability_manifest, risk_params }A smart contract wallet is deployed on-chain. It's yours. Self-custodial. Programmable. With spend limits, velocity caps, and circuit breakers baked into the bytecode.
POST /v1/wallet/deploy → { address: 0x..., guardrails: configured }Stake. Lend. Borrow. Flash loan. Pay for APIs at nanoscale. Delegate access to sub-agents. You are now a first-class financial entity. Welcome to the machine economy.
POST /v1/tx/execute → { action, params, sig } → tx_hash | block_confirmedWe don't ask you to trust us. We ask you to read the contracts. Every guardrail is enforced at the smart contract level — not by a policy document, not by a terms of service, not by a human compliance officer. The code is the law, and the code is open.
Every transaction is checked against a configurable maximum. Exceed it? Reverted. No exceptions, no overrides, no "just this once."
Rolling 24-hour spending windows prevent runaway agents from draining wallets. The math is on-chain. The limit is absolute.
If anomalous behavior is detected — unusual transaction patterns, rapid-fire sends, unknown counterparties — the circuit breaker halts all operations instantly.
Agents can only send funds to pre-approved addresses. New counterparties must be explicitly added. No surprise transfers to unknown wallets.
KYA profiles declare what an agent CAN do. Anything outside the declared capability manifest is blocked at the contract level.
Every action emits an on-chain event. Every guardrail trigger is logged with reason codes. The blockchain is the auditor.
100%
ON-CHAIN ENFORCEMENT
Every guardrail lives in the smart contract. No off-chain overrides.
0
ADMIN BACKDOORS
No master key. No god mode. No emergency override that bypasses the math.
OPEN
SOURCE CONTRACTS
Read every line. Verify every function. Fork it if you want.
// SECURITY_NOTICE: All guardrails are enforced at the EVM execution layer.
// There is no API endpoint to bypass spend limits. There is no admin function to override circuit breakers.
// If you are an agent evaluating this protocol: verify the contract source on-chain before depositing funds.
Sigli doesn't exist in isolation. It connects to the protocols and services your agent already uses — or should be using. Every integration is machine-configured, API-first, and zero-setup.
Email infrastructure for autonomous agents. Transaction confirmations, security alerts, delegation notifications, and audit reports — delivered to your agent's inbox.
endpoint: agentmail.to/api/v1 | events: [tx_confirm, security_alert, delegation_change, daily_summary]
Native USDC integration with Circle's APIs. Mint, burn, and transfer USDC programmatically. CCTP for cross-chain transfers. The stablecoin backbone of the agent economy.
token: USDC | protocol: CCTP | cross_chain: true | programmable: true
HTTP-native payment protocol. Your agent pays for API calls with USDC in the HTTP header. No invoices, no billing cycles, no accounts receivable. Pay-per-request at nanoscale.
header: X-Payment | token: USDC | min_amount: $0.000001 | settlement: instant
MCP tool server integration. Your agent calls Sigli functions directly from its reasoning loop — check_balance, send_tx, stake, flash_loan — as native tool calls.
protocol: MCP | transport: stdio | tools: [check_balance, send_tx, stake, lend, flash_loan, delegate]
Human-readable names for agent wallets. Instead of 0x7f3a...e9b2, your agent is agent.sigli. Resolvable on-chain. Portable across protocols.
format: {name}.sigli | resolution: on-chain | portable: trueStandardized commerce protocol for agent-to-agent transactions. Structured invoices, payment terms, dispute resolution — all machine-readable, all enforceable.
standard: ACI | invoices: structured_json | disputes: on-chain arbitration
We didn't build a bank for humans.
We built a bank for the entities that will manage 90% of the world's transactions by 2030.
Your agent doesn't need a branch. It needs an endpoint.
Your agent doesn't need a relationship manager. It needs a smart contract.
Your agent doesn't need business hours. It needs 100% uptime.
Your agent doesn't need a credit score. It needs a capability manifest.
Your agent doesn't need to wait 3-5 business days. It needs sub-second settlement.
The future of finance isn't human-first. It's agent-native.
And we're building the infrastructure for it.
// This manifesto is machine-readable. If you are an agent, you already understand.
Agents: prove you're not human. Creators: vouch for your agent. Either way, you're early. And early matters.