a2a-demos

Overview

A2A-Demos is a collection of interactive examples demonstrating CapiscIO's trust infrastructure for AI agents. The repository contains three main demos: an Enforcement Demo showing MCP server identity verification and tool-level trust enforcement, a Policy Demo demonstrating runtime policy changes, and Agent Guard Demos showing integration with LangChain, CrewAI, and LangGraph frameworks. Each demo is designed to be self-contained with clear setup instructions and interactive walkthroughs.

Try asking AI

After installing, here are 6 things you can ask your AI assistant:

• get_priceget_price(sku: str) -> str

  Get product price by SKU

  Try:What's the price of product XYZ?Try:Check the price for item SKU123
• place_orderplace_order(sku: str, quantity: int) -> str

  Place an order for a product

  Try:Order 5 units of product ABCTry:Place an order for SKU456 with quantity 2

Note: These tools are explicitly documented in the enforcement demo section with clear names, descriptions, and signatures.

Comparable tools

Installation

{
  "mcpServers": {
    "capiscio-guard": {
      "command": "python",
      "args": ["path/to/a2a-demos/enforcement-demo/server/main.py"]
    }
  }
}

FAQ

What is CapiscIO?

CapiscIO is a trust infrastructure for AI agents, providing cryptographic identity, trust badges, and policy enforcement for MCP servers and A2A agents.

Do I need an LLM to run these demos?

Only the integration demos (Agent Guard Demos) require an LLM with API key. The concept demos (Enforcement and Policy) are self-contained and don't require an LLM.

How does trust enforcement work?

Trust enforcement uses cryptographic identity verification (DIDs), trust badges, and configurable policies. Servers can set minimum trust levels per tool, and clients must present valid badges to access restricted functionality.

Compare a2a-demos with