Welcome to langchain-modexia, the premium provider package for AI Developers building intelligent, financially autonomous agents in Python.

This package provides native LangChain tools allowing your agents to instantly hold USDC, execute cross-chain transfers, open zero-gas micro-payment channels, and automatically negotiate HTTP paywalls without needing to handle private keys directly in the LLM context.

Installation

Install the package directly via pip. This will also automatically install the core cryptographic engine (modexiaagentpay) under the hood.

pip install langchain-modexia

Quick Start (LangGraph / Agents)

For the absolute smoothest developer experience, we provide the ModexiaToolkit. This automatically injects the entire suite of 8+ crypto capabilities seamlessly into standard LangGraph or AgentExecutor workflows.

1. Synchronous Agents (Standard)

import os
from langchain_modexia import ModexiaToolkit
from langchain_anthropic import ChatAnthropic
from langgraph.prebuilt import create_react_agent

# Use the API key generated from https://modexia.software
os.environ["MODEXIA_API_KEY"] = "mx_test_your_key_here_..."

# Initialize the toolkit securely
toolkit = ModexiaToolkit.from_api_key(os.environ["MODEXIA_API_KEY"])
tools = toolkit.get_tools()

# Attach tools to your favorite LLM
llm = ChatAnthropic(model="claude-3-5-sonnet-20240620")
agent_executor = create_react_agent(llm, tools)

# The Agent is ready to execute Modexia payments!
response = agent_executor.invoke({
    "messages": [("user", "Check my wallet balance and send 5 USDC to 0x123abc...")]
})
print(response["messages"][-1].content)

2. High-Concurrency Asynchronous Swarms (Async)

If you are running multi-agent swarms (e.g. ainvoke), Modexia flawlessly supports fully non-blocking I/O. Just pass async_mode=True when creating the toolkit!

import asyncio

async def main():
    # Setting async_mode=True automatically provisions the AsyncModexiaClient
    # and routes LangChain 'ainvoke' to non-blocking HTTPx calls.
    toolkit = ModexiaToolkit.from_api_key(os.environ["MODEXIA_API_KEY"], async_mode=True)
    tools = toolkit.get_tools()
    
    agent_executor = create_react_agent(llm, tools)
    
    res = await agent_executor.ainvoke({
       "messages": [("user", "Transfer 1.5 USDC to Alice on Arbitrum using CCTP")]
    })

asyncio.run(main())

Native Tools Injected

The ModexiaToolkit automatically handles schema validation for the LLM using strict Pydantic models. The following tools are injected:

• modexia_get_balance: Checks the Agent's current usable USDC balance.
• modexia_transfer: Sends a standard on-chain Modexia transaction to any EVM compatible address.
• modexia_cross_chain_transfer: Routes and burns USDC natively across any blockchain using Squid Router CCTP abstraction.
• modexia_get_history: Retrieves recent financial history for LLM context injection.
• modexia_smart_fetch: Allows the agent to hit an external URL and automatically pay HTTP 402 Payment Required L402 invoices to acquire premium data!
• modexia_open_channel: Opens an ongoing high-frequency payment vault.
• modexia_consume_channel: Executes an instant, gas-free cryptographically signed micro-payment inside an open channel (perfect for sub-cent API usage).
• modexia_settle_channel: Closes the vault and distributes funds.

🌐 JavaScript & Non-Python Environments

Are you building your LangChain App in TypeScript/Node.js, or using Claude Desktop?

Because Python agents get a dedicated native package (langchain-modexia), we recommend non-Python systems use our Model Context Protocol (MCP) server instead!

Simply install modexia-mcp and connect to it using @langchain/mcp-adapters to unlock identical functionality in JavaScript with zero custom code!

🤝 Contributing

Found a bug or want to add a feature? Open a Pull Request! This package acts as a thin LangChain abstraction over the core Modexia AgentPay Python SDK.