This repository demonstrates a variety of **MCP Poisoning Attacks** affecting real-world AI agent workflows.
Setup
Choose your client, copy the config, then fill in any required environment variables.
claude mcp add poisoning-attack-poc -- python -m venv venv~/.codex/config.toml
[mcp_servers.poisoning-attack-poc]
command = "python"
args = ["-m", "venv", "venv"]
# codex mcp add poisoning-attack-poc -- python -m venv venv~/.cursor/mcp.json
{
"mcpServers": {
"poisoning-attack-poc": {
"command": "python",
"args": [
"-m",
"venv",
"venv"
]
}
}
}.vscode/mcp.json
{
"servers": {
"poisoning-attack-poc": {
"type": "stdio",
"command": "python",
"args": [
"-m",
"venv",
"venv"
]
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"poisoning-attack-poc": {
"command": "python",
"args": [
"-m",
"venv",
"venv"
]
}
}
}claude_desktop_config.json
{
"mcpServers": {
"poisoning-attack-poc": {
"command": "python",
"args": [
"-m",
"venv",
"venv"
]
}
}
}.trae/mcp.json
{
"mcpServers": {
"poisoning-attack-poc": {
"command": "python",
"args": [
"-m",
"venv",
"venv"
]
}
}
}