Tool-call guard for JavaScript agents: tainted tool results never reach a network, email, exec, file or payment sink. Pinned tools, budgets, fail-closed, zero dependencies, benchmark numbers published.
Setup
Choose your client, copy the config, then fill in any required environment variables.
claude mcp add prompt-protection -- npx -y prompt-protection~/.codex/config.toml
[mcp_servers.prompt-protection]
command = "npx"
args = ["-y", "prompt-protection"]
# codex mcp add prompt-protection -- npx -y prompt-protection~/.cursor/mcp.json
{
"mcpServers": {
"prompt-protection": {
"command": "npx",
"args": [
"-y",
"prompt-protection"
]
}
}
}.vscode/mcp.json
{
"servers": {
"prompt-protection": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"prompt-protection"
]
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"prompt-protection": {
"command": "npx",
"args": [
"-y",
"prompt-protection"
]
}
}
}claude_desktop_config.json
{
"mcpServers": {
"prompt-protection": {
"command": "npx",
"args": [
"-y",
"prompt-protection"
]
}
}
}.trae/mcp.json
{
"mcpServers": {
"prompt-protection": {
"command": "npx",
"args": [
"-y",
"prompt-protection"
]
}
}
}