Skip to content

README.md

Latest commit

 

History

History
87 lines (61 loc) · 2.25 KB

README.md

File metadata and controls

87 lines (61 loc) · 2.25 KB

AVP Protocol Workshop

Hands-on demos and tutorials for the Agent Vault Protocol.

Overview

This workshop teaches you how to secure AI agent credentials using AVP. Each demo builds on the previous one, taking you from basic usage to hardware-secured deployments.

Prerequisites

  • Python 3.9+
  • An API key (Anthropic, OpenAI, or similar)
  • 15-30 minutes per demo

Demos

# Demo Description Difficulty
01 OpenClaw Basic Replace insecure keys.json with AVP Beginner
02 LangChain + AVP Secure LangChain credentials Beginner
03 CrewAI + AVP Multi-agent credential management Intermediate
04 Hardware Security NexusClaw hardware key integration Advanced

Quick Start

# Clone this repo
git clone https://github.com/avp-protocol/workshop.git
cd workshop

# Start with Demo 01
cd demos/01-openclaw-basic
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# Follow the README in each demo folder

What You'll Learn

  • Demo 01: Migrate from plaintext credentials to encrypted AVP storage
  • Demo 02: Integrate AVP with LangChain for secure LLM applications
  • Demo 03: Manage credentials across multiple CrewAI agents
  • Demo 04: Use hardware security keys for maximum protection

The Problem We're Solving

Most AI agent frameworks store API keys insecurely:

# Bad: Plaintext in environment
export ANTHROPIC_API_KEY=sk-ant-api03-...

# Bad: Plaintext in .env file
ANTHROPIC_API_KEY=sk-ant-api03-...

# Bad: Plaintext in keys.json
{"anthropic_api_key": "sk-ant-api03-..."}

These are trivially stolen by:

  • Infostealer malware
  • Compromised dependencies
  • Accidental git commits
  • Process memory dumps

The AVP Solution

# Good: Encrypted AVP vault
import avp

vault = avp.Vault("avp.toml")
api_key = vault.retrieve("anthropic_api_key")
# Key is encrypted at rest, never touches disk in plaintext

Resources

License

Apache 2.0