AI safety middleware — detects self-harm and criminal intent in LLM prompts.
uvx humane-proxy{
"OPENAI_API_KEY": "YOUR_SECRET_VALUE",
"GROQ_API_KEY": "YOUR_SECRET_VALUE"
}Add this server entry to the mcpServers object in your Claude Desktop config, then restart the app.
{
"mcpServers": {
"io-github-vishisht16-humane-proxy": {
"command": "uvx",
"args": [
"humane-proxy"
],
"env": {
"OPENAI_API_KEY": "YOUR_SECRET_VALUE",
"GROQ_API_KEY": "YOUR_SECRET_VALUE"
}
}
}
}~/Library/Application Support/Claude/claude_desktop_config.json%APPDATA%\Claude\claude_desktop_config.jsonNo remote HTTP endpoint is advertised. Use the package or stdio setup shown in Install.
humane-proxy is an MCP server for AI safety middleware — detects self-harm and criminal intent in LLM prompts.. It supports STDIO transport.
Use the generated config in Install. This server runs with uvx humane-proxy; add any required environment variables before starting your client.
Choose the Claude Desktop tab in Install and copy the config for uvx humane-proxy. Add required environment variables before starting Claude Desktop.
Choose the Claude Code tab in Install and copy the config for uvx humane-proxy. Add required environment variables before starting Claude Code.
Choose the Codex tab in Install and copy the config for uvx humane-proxy. Add required environment variables before starting Codex.
Choose the Cursor or VS Code tab in Install and copy the config for uvx humane-proxy. Add required environment variables before starting Cursor or VS Code.
humane-proxy uses STDIO transport. Use the package or command config in Install.
humane-proxy inventory is listed when the MCP endpoint exposes tools, resources, or prompts. Some servers require auth first.
humane-proxy does not advertise a verified auth requirement. If discovery fails, it may still need provider login, an API key, a bearer token, or a session header.
| Package | Registry | Version | Inputs |
|---|---|---|---|
humane-proxystdio | pypi | 0.4.0 | Env: OPENAI_API_KEY secret Env: GROQ_API_KEY secret |