Humane Proxy runs on your own machine — the client starts it, so there is no endpoint to ping. 178 installs a week from pypi. Last commit 29 Jul 2026.
AI safety middleware — detects self-harm and criminal intent in LLM prompts.
Today is the operative word: we check Humane Proxy every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add humane-proxy -- uvx humane-proxy
{
"mcpServers": {
"humane-proxy": {
"args": [
"humane-proxy"
],
"command": "uvx"
}
}
}
[mcp_servers.humane-proxy]
command = "uvx"
args = ["humane-proxy"]
{
"mcpServers": {
"humane-proxy": {
"args": [
"humane-proxy"
],
"command": "uvx"
}
}
}
{
"mcpServers": {
"humane-proxy": {
"args": [
"humane-proxy"
],
"command": "uvx"
}
}
}
This one needs environment variables set before it will start:
OPENAI_API_KEY (OpenAI API key for Stage 3 reasoning (optional)), GROQ_API_KEY (Groq API key for LlamaGuard Stage 3 (optional)).
The author declared them in the registry entry; get the values from the project itself.
Multi-agent LLM security layer detecting prompt injection and jailbreaks.
Extracts decision intent from git history and protects intentional code from AI modification.
Check an email before an AI agent sends it: send / ask a human / block. Detects prompt injection.
Local-first PHI detection for source code, LLM prompts, logs, and analytics
Detect prompt injection, jailbreak, and social-engineering attacks in LLM agents.
Safety-aware MCP server for Linear issues, projects, cycles, and dependencies.
Prompt injection + PII detection for LLM apps and agent pipelines. Deterministic, sub-25ms.
Security middleware for MCP. Blocks prompt injection, PII leakage, and resource exhaustion.
Answers built from our own checks of this server.