Finds every spot where LLM output reaches a user or triggers a side-effect — and flags the unguarded ones. CLI + MCP.
Setup
Choose your client, copy the config, then fill in any required environment variables.
claude mcp add aiglare -- npx -y @nugehs/aiglare~/.codex/config.toml
[mcp_servers.aiglare]
command = "npx"
args = ["-y", "@nugehs/aiglare"]
# codex mcp add aiglare -- npx -y @nugehs/aiglare~/.cursor/mcp.json
{
"mcpServers": {
"aiglare": {
"command": "npx",
"args": [
"-y",
"@nugehs/aiglare"
]
}
}
}.vscode/mcp.json
{
"servers": {
"aiglare": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@nugehs/aiglare"
]
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"aiglare": {
"command": "npx",
"args": [
"-y",
"@nugehs/aiglare"
]
}
}
}claude_desktop_config.json
{
"mcpServers": {
"aiglare": {
"command": "npx",
"args": [
"-y",
"@nugehs/aiglare"
]
}
}
}.trae/mcp.json
{
"mcpServers": {
"aiglare": {
"command": "npx",
"args": [
"-y",
"@nugehs/aiglare"
]
}
}
}