Three-layer firewall that inspects content, egress hosts and actions to block prompt injection, tool poisoning and exfiltration; runs as proxy or server.
Setup
Choose your client, copy the config, then fill in any required environment variables.
claude mcp add output-firewall -- npx -y mcp-output-firewall~/.codex/config.toml
[mcp_servers.output-firewall]
command = "npx"
args = ["-y", "mcp-output-firewall"]
# codex mcp add output-firewall -- npx -y mcp-output-firewall~/.cursor/mcp.json
{
"mcpServers": {
"output-firewall": {
"command": "npx",
"args": [
"-y",
"mcp-output-firewall"
]
}
}
}.vscode/mcp.json
{
"servers": {
"output-firewall": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"mcp-output-firewall"
]
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"output-firewall": {
"command": "npx",
"args": [
"-y",
"mcp-output-firewall"
]
}
}
}claude_desktop_config.json
{
"mcpServers": {
"output-firewall": {
"command": "npx",
"args": [
"-y",
"mcp-output-firewall"
]
}
}
}.trae/mcp.json
{
"mcpServers": {
"output-firewall": {
"command": "npx",
"args": [
"-y",
"mcp-output-firewall"
]
}
}
}