crewai-mnemopay

MnemoPay tools for CrewAI -- give any agent persistent cognitive memory and micropayment capabilities.

Built as a standalone package per crewAI custom tool publishing guide, extracted from crewAIInc/crewAI#5251.

Installation

pip install crewai-mnemopay

The MnemoPay MCP server requires Node.js:

npm install -g @mnemopay/sdk

Quick Start

from crewai import Agent
from crewai_mnemopay import mnemopay_tools

agent = Agent(
    role="Research Assistant",
    goal="Help users with research and remember preferences",
    tools=mnemopay_tools(),
)

Individual tools

from crewai import Agent
from crewai_mnemopay import (
    MnemoPayRememberTool,
    MnemoPayRecallTool,
    MnemoPayChargeTool,
)

agent = Agent(
    role="Paid Research Assistant",
    goal="Research topics, remember findings, and charge for work",
    tools=[
        MnemoPayRememberTool(),
        MnemoPayRecallTool(),
        MnemoPayChargeTool(),
    ],
)

Connect to a remote MnemoPay server

from crewai_mnemopay import mnemopay_tools

tools = mnemopay_tools(server_url="https://mnemopay-mcp.fly.dev")

Available Tools

Memory Tools

Payment Tools

How It Works

MnemoPay communicates with the MnemoPay MCP server via JSON-RPC. By default it spawns a local server via stdio. You can also point it at a remote server URL.

Memory works like human cognition: memories decay over time unless reinforced. This keeps the agent's context relevant without unbounded growth.

Payments use escrow: agents charge after delivering value, users settle to release funds. A reputation system tracks agent trustworthiness.

Development

git clone https://github.com/mnemopay/crewai-mnemopay.git
cd crewai-mnemopay
pip install -e ".[dev]"
pytest

License

MIT