Trusted MCP

Trusted MCP adds component attestation and verification on top of the standard MCP protocol, plus end-to-end encrypted communication. Combined with confidential computing (such as Jeddak AICC) and trusted inference services, it guards against untrusted service identities, tampered data, traffic hijacking, and privacy leaks.

Trusted agents

With confidential computing, you can run the agent and its model services (such as confidential Doubao) in a trusted environment, then use Trusted MCP’s end-to-end encrypted communication to build a fully trusted agent. VeADK natively supports connecting to Trusted MCP services, opening a trusted channel within the tool-call chain to secure outbound communication.

Prerequisites

Prepare a Trusted MCP service

Obtain a Trusted MCP service address, e.g. wss://your-trusted-mcp.example.com.

Prepare the AICC config

Prepare the AICC config file ./aicc_config.json (for the confidential environment and remote attestation); see the official example.

Usage

Turn on x-trusted-mcp: true in the connection parameters, then connect with TrustedMcpToolset:

agent.py

import asyncio

from veadk import Agent
from veadk.utils.mcp_utils import get_mcp_params
from veadk.tools.mcp_tool.trusted_mcp_toolset import TrustedMcpToolset

mcp_url = "<Trusted MCP server address>"

# Open the trusted channel
connection_params = get_mcp_params(mcp_url)
connection_params.headers = {"x-trusted-mcp": "true"}

# Initialize the trusted toolset and attach it to the agent
toolset = TrustedMcpToolset(connection_params=connection_params)
agent = Agent(tools=[toolset])

response = asyncio.run(agent.run("What's the weather in Beijing?"))
print(response)

Options

Key TrustedMcpToolset options:

Option	Location	Default	Description
`x-trusted-mcp`	Header	`true`	Whether to enable Trusted MCP and open the trusted channel
`aicc_config`	File path	`./aicc_config.json`	AICC config file path, for attestation and policy in the confidential-computing environment

For the config file’s format and details, see the Trusted MCP config reference.

Overview

Get started

Core components

Observability

Extensions

Frontend

Security

Trusted agents

Prerequisites

Usage

Options

​Trusted agents

​Prerequisites

​Usage

​Options

Trusted agents

Prerequisites

Usage

Options