Developer Documentation

AI Agent Integration

stillrunning is the only security tool built for the AI coding era. Whether you're using Claude Code, Cursor, or Devin — stillrunning checks every package your AI installs.

Full setup instructions for all AI agents →

Install Protection

Block malicious packages at install time.

# Install stillrunning pip install stillrunning # Run setup wizard (3 minutes) stillrunning --setup

The setup wizard configures pip/npm interception and connects to the live threat feed.

Scan a package from the CLI

# Scan a PyPI package stillrunning scan requests # Scan an npm package (v2.13.2+, scoped @org/pkg supported) stillrunning scan @types/node --npm

Import Protection

Block malicious packages at runtime, even if they sneak past install checks.

# Add to top of your main script import stillrunning.hook # Any malicious import will raise ImportError

Always-on Mode

# Install .pth file for automatic protection stillrunning --install-hook # Remove if needed stillrunning --uninstall-hook

MCP / Claude Code Integration

Let Claude check packages before installing them.

{
  "mcpServers": {
    "stillrunning": {
      "type": "url",
      "url": "https://stillrunning.io/mcp",
      "name": "stillrunning"
    }
  }
}

When you ask Claude to install a package, it automatically checks stillrunning first.

API Reference

Check Package

curl "https://stillrunning.io/api/check-package?name=requests&version=2.31.0" # Response: # {"package": "requests", "verdict": "CLEAN", "score": 5, "cached": true, ...}

Scan Package (AI tier)

curl -X POST "https://stillrunning.io/api/scan" \ -H "Content-Type: application/json" \ -d '{"token": "YOUR_TOKEN", "package": "requests", "ecosystem": "pip"}'

MCP Endpoint

curl -X POST "https://stillrunning.io/mcp" \ -H "Authorization: Bearer YOUR_TOKEN" \ -H "Content-Type: application/json" \ -d '{"method": "tools/call", "params": {"name": "check_package", "arguments": {"package_name": "requests"}}}'

Interactive Approvals

Unknown packages trigger Telegram alerts with one-tap approve/deny:

CLEAN (score <50): Allowed silently
UNKNOWN (not seen): Blocked pending approval (60s timeout)
SUSPICIOUS (score 50-79): Blocked pending approval
DANGEROUS (score 80+): Hard block, no override

Rate Limits

Tier	Package scans/day	AI scans/day
Free	10	0
Personal ($9)	100	0
Basic ($29)	Unlimited	0
AI ($49)	Unlimited	100
Enterprise ($499)	Unlimited	10,000

See full coverage details →