Requirements

import { Aside } from ‘@astrojs/starlight/components’;

System

Requirement	Minimum version	Notes
Python	3.11+	Tested on 3.11 and 3.12
Operating system	macOS, Linux, Windows (WSL2)	Native Windows not recommended
RAM	8 GB minimum	16 GB recommended for Qwen3:14b
Storage	10 GB free	For Ollama models and data

Ollama (for local mode)

Ollama is required if you want to run the ReAct agent completely locally, without depending on external LLM APIs.

# macOS
brew install ollama

# Linux
curl -fsSL https://ollama.ai/install.sh | sh

After installing, pull a model with tool calling support:

# Recommended (16 GB RAM) — best quality/speed balance
ollama pull qwen3:14b

# If you have only 8-16 GB RAM
ollama pull qwen3:8b

# If you have 48 GB RAM — maximum quality
ollama pull qwen3:32b

LLM Alternatives

If you don’t want to install Ollama, you can use:

OpenAI (gpt-4o) — requires OPENAI_API_KEY
Anthropic (claude-3-5-sonnet-20241022) — requires ANTHROPIC_API_KEY

With these options the MCP Server works without Ollama.

OSINT API Keys (optional)

Most tools have a free tier. You only need at least one to get started. VirusTotal is recommended.

Service	Variable	Free tier	Register
VirusTotal	`VIRUSTOTAL_API_KEY`	500 req/day	virustotal.com
AbuseIPDB	`ABUSEIPDB_API_KEY`	1,000 req/day	abuseipdb.com
AlienVault OTX	`ALIENVAULT_OTX_API_KEY`	Unlimited	otx.alienvault.com
Shodan	`SHODAN_API_KEY`	100 req/month	shodan.io
IPInfo	`IPINFO_TOKEN`	50,000 req/month	ipinfo.io
URLScan.io	`URLSCAN_API_KEY`	5,000 req/day	urlscan.io