modexiaagentpay 0.7.0

Modexia AgentPay — Python SDK for agent wallets & payments (USDC)

---

noteId: "8d854f304bd411f18bfe97159fe8ca5e" tags: []

---

🚀 Modexia Python SDK

The ultimate Python client for AI Agents to seamlessly interact with Modexia's AgentPay API.

Welcome to the Modexia Python SDK (modexiaagentpay). Built from the ground up to give your AI agents and Python applications frictionless access to Modexia's wallet and payment infrastructure. Enable your AI to hold, manage, and transfer value with just a few lines of code!

---

📋 What's New in v0.7.0

Intent-to-Pay — A new cryptographically signed payment system that gives agents rich compliance feedback, audit trails, and policy-aware spending.

Feature	Description
client.pay()	High-level intent-based payment — create, sign, submit, and poll in one call
client.create_intent()	Create an HMAC-SHA256 signed intent token
client.submit_intent()	Submit a signed token through the 11-step validation pipeline
client.get_intent()	Look up the status of a previously submitted intent
client.list_intents()	List recent intents for audit trail review
IntentResult	New dataclass with status, compliance metadata, daily spend, and suggestions
memo parameter	All payment methods now accept a memo for audit trail visibility

---

🌟 Getting Started: Your API Key

Before writing your first integration, you will need a Modexia developer account and an API key.

1. Visit modexia.software
2. Create or log into your developer account.
3. Navigate to your dashboard and generate your API Key.

For deep dives, architecture, and advanced agentic payment flows, dive into our Full Documentation.

---

🏗 System Architecture & Flow

The SDK supports two payment flows: the classic v1 transfer and the new v2 intent-based pipeline.

Intent-to-Pay Flow (v2 — Recommended)

sequenceDiagram
    participant Agent as 🤖 AI Agent
    participant SDK as 📦 Python SDK
    participant API as 🌐 Modexia API
    participant Pipeline as 🔒 Validation Pipeline
    participant Chain as ⛓️ Blockchain

    Agent->>SDK: client.pay(recipient, amount, memo="...")
    SDK->>SDK: Create intent JSON + HMAC-SHA256 sign
    SDK->>API: POST /v2/intents/submit (signed token)
    API->>API: Verify HMAC signature
    API->>Pipeline: Run 11-step validation
    
    Note over Pipeline: amount ✓ recipient ✓ expiry ✓<br/>nonce ✓ wallet ✓ KYC ✓<br/>SBT ✓ policy ✓ limits ✓

    alt All checks pass
        Pipeline-->>API: approved
        API->>Chain: Execute on-chain transfer
        Chain-->>API: txId
        API-->>SDK: IntentResult(status="executed", txId, daily_remaining, balance)
        SDK-->>Agent: Rich result with compliance metadata 🎉
    else Validation fails
        Pipeline-->>API: rejected (code + suggestion)
        API-->>SDK: IntentResult(status="rejected", reason, suggestion)
        SDK-->>Agent: Actionable rejection with fix suggestion
    end

Classic Transfer Flow (v1 — Still Supported)

sequenceDiagram
    participant Agent as 🤖 AI Agent
    participant SDK as 📦 Python SDK
    participant API as 🌐 Modexia API
    participant Chain as ⛓️ Blockchain

    Agent->>SDK: client.transfer(recipient, amount)
    SDK->>API: POST /v1/agent/pay
    Note over API: Internally creates synthetic intent<br/>and runs same validation pipeline
    API->>Chain: Execute transfer
    Chain-->>API: txId
    API-->>SDK: {success: true, txId}
    SDK-->>Agent: PaymentReceipt 🎉

Note: The v1 transfer() method now internally routes through the same intent validation pipeline. You get the same compliance enforcement with zero code changes.

---

✨ Top-Level Features

• Intent-Based Payments: Cryptographically signed payment intents with HMAC-SHA256 verification, replay protection, and full compliance pipeline.
• Rich Compliance Feedback: Know exactly why a payment was rejected, with actionable suggestions and your daily spend/remaining budget.
• Audit Trail: Every payment records a memo explaining why the AI made the payment — visible in the dashboard.
• Built for AI Agents: Designed specifically for programmatic access to agent wallets and payments without complex cryptography.
• Async & "Swarm" Support: High-concurrency operations powered by blazing fast implementations in both synchronous and asyncio models.
• Fully Typed: Returns robust Python dataclasses for exceptional developer experience and editor autocompletion.
• Reliable Networking: Built-in retry and exponential backoff mechanisms to gracefully handle transient network errors.

---

📦 Installation

Install the production-ready package directly from PyPI:

pip install modexiaagentpay

For local development and contribution:

git clone https://github.com/Modaniel/SDKs.git
cd SDKs
pip install -e packages/SDKs/pythonSdk

---

🚀 Quick Start Guide

Intent-Based Payment (v2 — Recommended)

The pay() method is the new recommended way to make payments. It creates a signed intent, validates it through the compliance pipeline, and returns rich metadata.

from modexia import ModexiaClient

client = ModexiaClient(api_key="mx_test_your_api_key_here")

# Make a payment with full compliance feedback
result = client.pay(
    recipient="0xabc123456789def0123456789abc0123456789def",
    amount=5.0,
    memo="Paying for GPT-4 API call to summarize document #42"
)

if result.status == "executed":
    print(f"✅ Payment executed! TX ID: {result.txId}")
    print(f"💰 Balance after: {result.wallet_balance_after} USDC")
    print(f"📊 Daily remaining: {result.daily_remaining} USDC")
elif result.status == "rejected":
    print(f"❌ Rejected: {result.reason}")
    print(f"💡 Suggestion: {result.suggestion}")

Classic Transfer (v1 — Backwards Compatible)

from modexia import ModexiaClient

client = ModexiaClient(api_key="mx_test_your_api_key_here")

# Check balance
balance = client.retrieve_balance()
print(f"💰 Balance: {balance} USDC")

# Simple transfer (still works, now with memo support)
receipt = client.transfer(
    recipient="0xabc123456789def0123456789abc0123456789def",
    amount=5.0,
    wait=True,
    memo="Paying vendor for compute resources"
)

if receipt.success:
    print(f"✅ Transfer successful! TX ID: {receipt.txId}")

Async Usage (Swarm Mode)

For high-concurrency loops or AutoGPT-style agent networks:

import asyncio
from modexia import AsyncModexiaClient

async def main():
    async with AsyncModexiaClient(api_key="mx_test_your_api_key_here") as client:
        await client.validate_session()

        # Intent-based payment (recommended)
        result = await client.pay(
            "0xabc...", 5.0,
            memo="Async swarm payment for data processing"
        )
        print(f"Status: {result.status}, Daily remaining: {result.daily_remaining}")

        # Or classic transfer
        receipt = await client.transfer("0xabc...", 2.0, memo="Legacy flow")
        print(f"TX: {receipt.txId}")

asyncio.run(main())

---

🛠 API Reference

ModexiaClient & AsyncModexiaClient

The core classes for all network operations.

ModexiaClient(
    api_key: str, 
    timeout: int = 15, 
    base_url: Optional[str] = None, 
    validate: bool = True
)

Intent-to-Pay Methods (v2)

Method	Description	Returns
pay(recipient, amount, memo=None, wait=True)	Recommended. High-level intent-based payment — creates, signs, submits, and polls. Returns rich compliance metadata.	IntentResult
create_intent(recipient, amount, memo=None, action="transfer", ttl_seconds=300)	Create a signed intent token locally (no network call). The token is base64url(payload).hex(hmac_sha256(payload, api_key)).	str
submit_intent(intent_token)	Submit a signed intent token for validation and execution.	IntentResult
get_intent(intent_id)	Retrieve the status of a previously submitted intent by UUID.	IntentResult
list_intents(limit=20)	List recent payment intents for audit trail review.	List[IntentResult]

Classic Methods (v1)

Method	Description	Returns
retrieve_balance() / get_balance()	Fetches the current available balance of your agent's wallet.	str
transfer(recipient, amount, idempotency_key=None, wait=True, memo=None)	Send funds to a destination. Uses intent-hashing to prevent duplicate charges.	PaymentReceipt
cross_chain_transfer(to_chain, to_token, recipient, amount)	Cross-chain CCTP transfer via Squid Router.	PaymentReceipt
get_history(limit=5)	Fetch the recent transaction history for Agent memory. Now includes memo.	TransactionHistoryResponse

Vault / Micropayment Channel Methods

Method	Description	Returns
open_channel(provider, deposit, duration_hours=24)	Open a payment channel with on-chain deposit.	dict
consume_channel(channel_id, amount, idempotency_key=None)	Execute an instant, gas-free micro-payment inside a channel.	ConsumeResponse
settle_channel(channel_id)	Settle a channel on-chain — pays provider, refunds remainder.	dict
get_channel(channel_id)	Get the current status of a payment channel.	ChannelStatus
list_channels(limit=50)	List all payment channels for the authenticated agent.	List[ChannelStatus]

Other

Method	Description	Returns
smart_fetch(method, url, **kwargs)	Auto-negotiates 402 HTTP paywalls automatically.	Response

---

📊 Data Models

IntentResult (New in v0.7.0)

The rich response returned from the intent-to-pay pipeline:

@dataclass
class IntentResult:
    status: str                          # pending | approved | rejected | executed | failed
    intent_id: Optional[str]             # UUID for tracking
    # Transaction details (populated on 'executed')
    txId: Optional[str]                  # Circle transaction ID
    txIds: Optional[List[str]]           # Multiple IDs for multi-leg txs
    txState: Optional[str]               # PENDING → COMPLETE
    amount: Optional[str]                # e.g. "5.00"
    recipient: Optional[str]             # 0x... address
    # Compliance & policy metadata
    wallet_balance_after: Optional[str]  # Balance after payment
    daily_spent: Optional[str]           # How much spent today
    daily_remaining: Optional[str]       # Budget remaining today
    # Rejection info
    reason: Optional[str]                # Why it was rejected
    code: Optional[str]                  # Machine-readable code
    suggestion: Optional[str]            # Actionable fix suggestion
    # Full validation pipeline results
    validation: Dict[str, Any]           # Per-step validation outcomes

Example validation field:

{
    "amount": "valid",
    "recipient": "valid",
    "expiry": "valid",
    "nonce": "valid",
    "wallet": "valid",
    "kyc": "valid",
    "sbt": "valid",
    "policy_registration": "valid",
    "policy_per_request": "valid",
    "policy_hourly": "valid",
    "policy_daily": "valid"
}

PaymentReceipt

@dataclass
class PaymentReceipt:
    success: bool
    status: str              # "PENDING" | "COMPLETE"
    txId: Optional[str]
    txHash: Optional[str]
    errorReason: Optional[str]

TransactionHistoryItem

@dataclass
class TransactionHistoryItem:
    txId: str
    type: str
    amount: str
    state: str
    createdAt: str
    providerAddress: Optional[str]
    txHash: Optional[str]
    memo: Optional[str]      # NEW in v0.7.0

---

🛑 Exception Handling

The SDK provides robust error mapping to help your agent gracefully recover from failures. You can import these directly from the modexia package:

from modexia import ModexiaClient, ModexiaAuthError, ModexiaPaymentError, ModexiaNetworkError

graph TD;
    Exception-->TimeoutError["TimeoutError: wait=True Polling exceeded 30s"];
    Exception-->ModexiaError;
    ModexiaError-->ModexiaAuthError["ModexiaAuthError: Key/Auth Issues"];
    ModexiaError-->ModexiaPaymentError["ModexiaPaymentError: Insufficient Funds / API Responses with success=False"];
    ModexiaError-->ModexiaNetworkError["ModexiaNetworkError: Connection Drops / Invalid JSON payloads"];

---

🔐 Intent Token Format

The intent token is a compact, tamper-proof string:

base64url(canonical_json).hex(hmac_sha256(canonical_json, api_key))

Canonical JSON is the payload with keys sorted alphabetically and serialized with no whitespace:

{"action":"transfer","amount":"5.0","currency":"USDC","expiresAt":1715270400000,"idempotencyKey":"uuid","memo":"Paying for API call","nonce":1715270100000,"recipient":"0x..."}

The HMAC-SHA256 signature uses your API key as the secret. The backend verifies this signature using constant-time comparison to prevent timing attacks.

---

🔄 Migration from v0.6.x

v0.7.0 is fully backwards compatible. No changes needed for existing code.

What's different internally: The transfer() method now routes through the same validation pipeline as pay(). Your existing code gets compliance enforcement for free.

To adopt v2 features:

- receipt = client.transfer("0x...", 5.0)
- if receipt.success:
-     print(f"TX: {receipt.txId}")

+ result = client.pay("0x...", 5.0, memo="Paying for compute")
+ if result.status == "executed":
+     print(f"TX: {result.txId}, Remaining: {result.daily_remaining}")

---

🤝 Contributing to Open Source

We actively welcome community contributions to make this SDK even better.

1. Fork & Branch: Open a Pull Request targeting the develop branch.
2. Backward Compatibility: Please keep API contracts stable. ModexiaClient, transfer(), and retrieve_balance() are canonical.
3. Testing: Make sure tests pass locally before submitting.

# Run tests from the repository root
pytest -q packages/SDKs/pythonSdk

---

📄 License & Support

modexiaagentpay is totally open-source and governed by the MIT License.

Need help scaling your agent swarm? Hit rate limits? Reach out to our engineering team or explore the docs at modexia.software.