Detect prompt injection, jailbreaks, and code injection in untrusted text before it reaches an LLM.
Setup
Choose your client, copy the config, then fill in any required environment variables.
claude mcp add safeprompt --env SAFEPROMPT_API_KEY='<SAFEPROMPT_API_KEY>' -- npx -y @safeprompt.dev/mcp~/.codex/config.toml
[mcp_servers.safeprompt]
command = "npx"
args = ["-y", "@safeprompt.dev/mcp"]
env = { SAFEPROMPT_API_KEY = "<SAFEPROMPT_API_KEY>" }
# codex mcp add safeprompt --env SAFEPROMPT_API_KEY='<SAFEPROMPT_API_KEY>' -- npx -y @safeprompt.dev/mcp~/.cursor/mcp.json
{
"mcpServers": {
"safeprompt": {
"command": "npx",
"args": [
"-y",
"@safeprompt.dev/mcp"
],
"env": {
"SAFEPROMPT_API_KEY": "<SAFEPROMPT_API_KEY>"
}
}
}
}.vscode/mcp.json
{
"servers": {
"safeprompt": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@safeprompt.dev/mcp"
],
"env": {
"SAFEPROMPT_API_KEY": "<SAFEPROMPT_API_KEY>"
}
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"safeprompt": {
"command": "npx",
"args": [
"-y",
"@safeprompt.dev/mcp"
],
"env": {
"SAFEPROMPT_API_KEY": "<SAFEPROMPT_API_KEY>"
}
}
}
}claude_desktop_config.json
{
"mcpServers": {
"safeprompt": {
"command": "npx",
"args": [
"-y",
"@safeprompt.dev/mcp"
],
"env": {
"SAFEPROMPT_API_KEY": "<SAFEPROMPT_API_KEY>"
}
}
}
}.trae/mcp.json
{
"mcpServers": {
"safeprompt": {
"command": "npx",
"args": [
"-y",
"@safeprompt.dev/mcp"
],
"env": {
"SAFEPROMPT_API_KEY": "<SAFEPROMPT_API_KEY>"
}
}
}
}Environment variables
SAFEPROMPT_API_KEY
Replace the placeholders with your own credentials and keep them out of version control.