Integrates with MalwareBazaar (a service by abuse.ch) to provide real-time threat intelligence, including querying recent malware samples, retrieving metadata by hash, and searching by tags.
Enables sending alerts and findings directly to a Telegram channel via a bot, supporting automated threat intelligence dissemination.
Integrates with VirusTotal to pull analysis statistics, reputation data, and file types for samples, with built-in API rate limiting to protect free-tier keys.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Malware Analysis MCP Server fetch the latest malware samples from MalwareBazaar"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A professional, modular, and secure Model Context Protocol (MCP) server providing real-time threat intelligence and malware sample metadata for AI assistants (like Claude, Cursor, and Windsurf).
It integrates MalwareBazaar, VirusTotal, and Telegram, along with 10+ offline analysis and crypto utilities, enabling AI models to autonomously aid in defensive malware research.
🔒 Security First: This server provides metadata and IOCs only. The downloading of live malware binaries is permanently disabled by design. Path traversal protections ensure safe local file hashing.
Real-Time Threat Intel: Query MalwareBazaar for recent samples, info by hash, or tags.
VirusTotal Integration: Pull analysis stats, reputation, and file types with built-in API rate limiting.
Advanced IOC Extraction: Extract and defang/refang IPs, Domains, URLs, Hashes, and Emails from unstructured text.
YARA generation: Ask the AI to generate a template YARA rule based on a sample's exact signatures.
Offline Crypto Lab: AES-GCM encryption (PBKDF2 600k iterations), SHA256/MD5 hashing, Shannon entropy analysis, and Hexdumps.
Alerting: Dispatch findings directly to a Telegram channel.
Secure Local Hashing: safely compute the SHA256 of local files to check against threat databases, strictly chrooted to allowed directories.
Related MCP server: Google Threat Intelligence MCP Server
This server works with any standard MCP client. Below are instructions for building the project and deploying it either to Claude Desktop (Mac/Windows) or testing it visually on Linux/Ubuntu using the MCP Inspector.
Ensure you have Node.js (v18 or newer) installed on your machine.
# 1. Clone the repository
git clone https://github.com/SaeidSec/Malware-Analysis-MCP-server.git
cd Malware-Analysis-MCP-server
# 2. Install dependencies
npm install
# 3. Build the TypeScript files
npm run buildMB_AUTH_KEY (Required): Get your free Auth-Key at auth.abuse.ch.
VT_API_KEY (Optional): VirusTotal API for vt_lookup. The free tier works perfectly (the server handles the 4 requests/minute limit).
TG_BOT_TOKEN & TG_CHAT_ID (Optional): Telegram bot token and chat ID for the send_alert tool.
ALLOWED_SCAN_DIRS (Highly Recommended): A comma-separated list of absolute paths (e.g., /home/user/Downloads,/tmp/malware). This strictly limits the check_local_file tool to prevent the AI from arbitrarily hashing personal files.
Claude Desktop is currently the most popular way to use MCP servers.
Open your Claude Desktop config file:
Mac: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Add the server configuration, pointing to your local built dist/index.js:
{
"mcpServers": {
"malware-analysis-mcp": {
"command": "node",
"args": [
"/absolute/path/to/Malware-Analysis-MCP-server/dist/index.js"
],
"env": {
"MB_AUTH_KEY": "YOUR_ABUSECH_AUTH_KEY",
"VT_API_KEY": "YOUR_VT_API_KEY",
"ALLOWED_SCAN_DIRS": "/Users/name/Downloads,/tmp/malware"
}
}
}
}Restart Claude Desktop completely. You can now prompt Claude to use your new threat intelligence tools!
Since Claude Desktop does not have an official Linux client yet, the best way to deploy and test the server on Ubuntu is using Anthropic's official MCP Inspector.
Create your environment variable file from the example:
cp example.env .env
nano .env # Add your API keys hereRun the MCP server directly into the Inspector:
npx -y @modelcontextprotocol/inspector node dist/index.jsOpen the provided http://localhost:5173 link in your web browser and click Connect. You will be presented with a beautiful UI where you can visually test tools like get_recent or extract_iocs manually.
get_recent - Fetch latest MalwareBazaar entries
get_info - Fetch precise sample metadata by SHA256
get_taginfo - Fetch samples by tag (e.g., "TrickBot")
virustotal_lookup - VT V3 detection stats and reputation
suggest_analysis_blogs - Links to Malpedia, MITRE, and writeups
extract_iocs - Pulls domains, IPs, MS5/SHA256, Emails from text
bulk_hash_check - Check multiple hashes against MalwareBazaar at once
threat_summary - Generates a curated Markdown report for a specific sample
generate_yara_template - Creates a starting point for YARA rules based on metadata
defang_iocs / refang_iocs - Safe IOC sharing (e.g., 1[.]1[.]1[.]1)
check_local_file - Securely hash a local file and query VT/MB
Encoding: base64_encode, base64_decode, url_encode, url_decode, hex_encode, hex_decode
Compression: gzip_compress, gzip_decompress
Crypto: aes_gcm_encrypt, aes_gcm_decrypt, hash_text
Analysis: entropy (Shannon), hexdump_text
send_alert - Push message to Telegram (TG_BOT_TOKEN and TG_CHAT_ID env vars required)
No Live Binaries: The get_file tool is disabled. The server cannot be tricked into executing or saving payload binaries.
Path Traversal Protection: Requests to check_local_file are resolved, normalized, and strictly checked against ALLOWED_SCAN_DIRS.
Hardened Crypto: The AES implementation uses PBKDF2 with 600,000 iterations (OWASP 2023 standard) and 16-byte random salts.
Rate Limiting: Protects your free-tier VirusTotal API key from accidental exhaustion by looping LLMs.
Streaming Hashing: Memory-safe streaming is used for hashing local files, preventing gigabyte-scale OOM crashes.
# Install dependencies
npm install
# Run typechecker
npm run typecheck
# Build the project
npm run build
# Run in dev mode
npm run devSee CONTRIBUTING.md for full guidelines.
MIT License. See LICENSE for details.
Inputs: envelope (required), password (required).
Inputs: password (required), salt, text.
Inputs: base64.
Inputs: text.
Inputs: hashes (required).
Inputs: path (required), vt.
Inputs: text (required).
Inputs: text.
Inputs: text (required), defang_output.
Inputs: rule_name (required), hash, author, description.
Inputs: no named parameters.
Inputs: hash (required).
Inputs: limit, selector.
Inputs: tag (required), limit.
Inputs: text.
Inputs: base64.
Inputs: alg (required), text.
Inputs: no named parameters.
Inputs: hex.
Inputs: text, limitBytes.
Inputs: text.
Inputs: text (required).
Inputs: message (required), useMarkdown.
Inputs: hash, signature.
Inputs: hash (required).
Inputs: text.
Inputs: text.
Inputs: hash (required).
Inputs: hash (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 29 tools
Several tools have vague or overlapping purposes, especially get_recent, get_info, get_taginfo, virustotal_lookup, and vt_lookup. The lack of descriptions makes it hard to distinguish similar-sounding intel lookup and generic retrieval tools from one another.
Most names use snake_case, which is readable, but the set mixes verb_noun patterns with noun-only names like entropy and hexdump_text. Abbreviation inconsistency between virustotal_lookup and vt_lookup weakens predictability.
At 29 tools, the server is heavy for a malware analysis utility surface and includes apparent duplication. The encoding, compression, hashing, and crypto operations could likely be grouped or reduced.
The tool set broadly covers threat intel lookup, local file checks, encoding, hashing, entropy, IOC handling, YARA templating, and alerting. It lacks obvious dynamic analysis or sandbox detonation capabilities, but the core analysis utility surface is fairly comprehensive.
Real-time threat intel for AI agents: 890K+ IOCs incl. prompt-injection & AI-skill threats
Real-time CVE, exploit, and vulnerability intelligence for AI assistants (350K+ CVEs, 115K+ PoCs)
AI-powered threat intelligence, smart contract auditing, and cybersecurity OSINT.
Pay-per-call cybersecurity for AI agents: vuln scans, threat intel, compliance, code security.
