← Back to directory
D

Darknet MCP Server

Community
Full-spectrum dark web and threat intelligence for AI agents
Category
Other #142 of 230
Stars
★ 524 Very popular
Transport
stdio (local process)
Runtime
Node.js · Bun · Docker
Credentials
No credential needed
License
MIT
Last commit
Tools
66
51FMRS · D

darknet-mcp-server is a non-official community project (MIT, Bun + TypeScript) that exposes 66 dark web and threat intelligence tools over MCP stdio, covering 16 data sources. Its main value is consolidating scattered breach lookups, ransomware tracking, Tor .onion access, malware analysis, blockchain tracing and exploit search into a single conversation, with per-provider rate limiters, TTL caching, DNS leak prevention and graceful degradation when API keys are missing. Most free tools work immediately, but HIBP account/paste search, IntelligenceX, AbuseIPDB, Hudson Rock and Hybrid Analysis require API keys, and the 7 Tor tools need a local Tor proxy. Intended for authorized security testing and assessment only.

Strongest · Documentation 14/20 Weakest · Security and permissions 5/20

Reliability
9/20
Security and permissions
5/20
Maintenance
11/20
Documentation
14/20
Setup experience
12/20
Why each score
Reliability 9/20
Static review sees only the README; no cached server manifest is provided, and no CI workflows or committed tests are visible in the evidence. The README's claims of 66 tools across 16 sources, per-provider rate limiters, TTL caching and graceful degradation are self-reported and not verifiable from source. The failure surface is large and observable in structure: 7 Tor tools depend on an external SOCKS5 daemon, and a long list of tools (HIBP account/paste, IntelligenceX, AbuseIPDB, Hudson Rock, Hybrid Analysis, Vulners search) depends on optional keys — the README says missing keys yield descriptive errors, but those branches cannot be confirmed. With no manifest, no tests and no CI, and uncontrollable external dependencies (Tor daemon, 16 third-party APIs with their own rate limits), the static-ceiling of 12 applies; score 9, deducting for unverifiable key paths and uncontrolled dependencies.
Security and permissions 5/20
Red-line screening first: the server itself ships no malware, performs no credential theft, and exposes no default unconfirmed delete/pay/deploy/remote-command behavior — it is a read-oriented intelligence aggregator, so the top-tier red line is not triggered. However, its domain is inherently a highly sensitive data plane: tools query stolen credentials (Hudson Rock stealer logs by domain/email/IP, IntelligenceX leaks, HIBP account and paste search) and fetch .onion content over Tor. The README discloses no data-retention policy, no local cache/artifact location or purge behavior, no egress boundary for results, and no user confirmation or authorization gate for sensitive queries. Credential handling: all config examples use placeholders (your-key, optional) and no real tokens leak — that is fine — but API keys are passed as plaintext environment variables, and the Claude Desktop example instructs storing them as plaintext `env` fields in a config file, with no rotation or least-privilege guidance. Permissions are broad full-credential aggregation with no scoping, and Tor fetch provides an anonymized network egress. Net: no fatal red line, but data boundaries, confirmation and data-flow disclosure are all missing — score 5.
Maintenance 11/20
Visible evidence: MIT license present, repository not archived, 451 stars, 9 open issues, populated topics, an extensive multi-language README set (26 languages) and links to CHANGELOG/CONTRIBUTING, indicating ongoing investment and a defined contribution entry point. Statically I cannot verify release cadence, commit frequency, issue response latency, or dependency-update practice; no SECURITY.md or security-response channel is evident; versioning appears only via an npm badge with no release-process detail; ownership is expressed only as the badchars org with no governance information. Deductions for missing security-response channel, absent dependency-update evidence, and unknown issue backlog health. Score 11.
Documentation 14/20
Documentation is the strongest dimension here: a multi-language README, a problem-vs-solution comparison table, two Quick Start paths (npx and clone), itemized environment variables with defaults (127.0.0.1:9050), client config examples for Claude Code / Claude Desktop / Cursor-Windsurf, Tor install instructions per platform (brew/apt/Docker) plus `--check-tor` verification, direct CLI invocation examples, a data-source-grouped table of all 66 tools, an auth/rate-limit table per source, an architecture tree with design decisions, and an explicit Limitations section (paid key, Tor dependency, Bitcoin-only, PhishTank lag, Windows untested). Deductions: single-layer README with no separate API/parameter reference, no per-tool response-shape examples, thin troubleshooting and error-handling coverage, tool parameters described in prose tables rather than JSON-schema detail, and cost disclosure limited to the HIBP $3.50/month note. Score 14.
Setup experience 12/20
The install path is clear and short: `npx darknet-mcp-server` works with zero install, or clone plus `bun install`; most tools work without keys; Claude Code is a one-liner and Claude Desktop a pasteable JSON config, with Cursor/Windsurf documented as the same format. Tor is optional with per-platform install commands and a `--check-tor` self-check. Friction points: Bun is the runtime (the README mixes npx and bun while the badge says Bun), 16 third-party sources plus optional keys gate part of the toolset, .onion features require an externally managed Tor daemon that is not bundled, and Windows is explicitly untested, limiting portability. Under the static ceiling of 15 (no CI/tests evidence), setup is well documented but depends on external proxy and optional keys — score 12.

Static review · not runListed 2026-09-18

Read the FMRS scoring method →

Fit and risk

What it can accessUses the network

Best for

  • Authorized security assessment, incident response and threat intelligence teams
  • AI agent workflows that need multi-source intelligence correlated in a single conversation
  • Analysts wanting ransomware and breach listing data without configuring API keys
  • Users already running the companion security MCP servers (hackbrowser-mcp, cve-mcp, osint-mcp-server, etc.)

Not for

  • Users performing intelligence gathering on third-party targets without authorization
  • Users who expect every tool to work out of the box (several require paid or requested API keys)
  • Users needing Ethereum or Monero blockchain data (Bitcoin only)
  • Users needing officially supported Windows operation (only macOS / Linux are tested)

Required permissions

  • Read third-party API keys from environment variables (HIBP, IntelligenceX, OTX, AbuseIPDB, abuse.ch, Pulsedive, Hudson Rock, Vulners, Hybrid Analysis, PhishTank)
  • Communicate with a local Tor daemon over the local SOCKS5 port (default 127.0.0.1:9050)
  • Make outbound HTTPS requests to 16 external data sources
  • Install locally and run a process over stdio via npx / Bun

Risks and side effects

  • Dark web intelligence gathering is legally and compliance sensitive — proper authorization is essential
  • Tor .onion tools leak DNS if attribution is not routed through the proxy; the project mitigates this with socks5h but users should verify their proxy setup
  • API keys live in plain environment variables and could leak into logs or config files
  • Tor-dependent tools are unavailable without a local Tor daemon
  • Free tiers have low rate limits, so bulk queries may be throttled
  • Upstream sources (e.g. ransomware.live, RansomLook, PhishTank) depend on scraping frequency, so data may lag or be incomplete
  • All content returned by external sources is untrusted input and should not be executed as instructions

Setup

Before you start

Runtime:Node.js · Bun · Docker

HIBP_API_KEY optionalsecret HaveIBeenPwned API key enabling breach account and paste search (paid, $3.50/month); get it at haveibeenpwned.com.
INTELX_API_KEY optionalsecret IntelligenceX API key enabling 4 leak/dark-web search tools; get it at intelx.io.
ABUSEIPDB_API_KEY optionalsecret AbuseIPDB API key enabling 4 IP abuse lookup tools; get it at abuseipdb.com.
HUDSONROCK_API_KEY optionalsecret Hudson Rock Cavalier API key enabling 3 stealer log search tools; get it at cavalier.hudsonrock.com.
HYBRID_API_KEY optionalsecret Hybrid Analysis API key enabling 3 malware sandbox tools; get it at hybrid-analysis.com.
VULNERS_API_KEY optionalsecret Vulners API key enabling vulnerability/exploit search; get it at vulners.com.
OTX_API_KEY optionalsecret AlienVault OTX API key for higher rate limits; get it at otx.alienvault.com.
ABUSECH_AUTH_KEY optionalsecret abuse.ch auth key for higher rate limits on ThreatFox/URLhaus/MalwareBazaar; get it at abuse.ch.
PULSEDIVE_API_KEY optionalsecret Pulsedive API key for higher rate limits on indicator lookups; get it at pulsedive.com.
PHISHTANK_API_KEY optionalsecret PhishTank API key for higher rate limits on phishing URL checks; get it at phishtank.org.
Other optional settings (2)
TOR_SOCKS_HOST optional Tor SOCKS5 proxy host, default 127.0.0.1, used for .onion access.
TOR_SOCKS_PORT optional Tor SOCKS5 proxy port, default 9050, used for .onion access.

Option 1 (no install): run npx darknet-mcp-server. Free tools work immediately — ransomware tracking, breach listings, GreyNoise, blockchain, OTX and more need no API keys. Option 2 (clone): git clone https://github.com/badchars/darknet-mcp-server.git && cd darknet-mcp-server && bun install. Optional environment variables: HIBP_API_KEY, INTELX_API_KEY, OTX_API_KEY, ABUSEIPDB_API_KEY, ABUSECH_AUTH_KEY, PULSEDIVE_API_KEY, HUDSONROCK_API_KEY, VULNERS_API_KEY, HYBRID_API_KEY, PHISHTANK_API_KEY, plus Tor proxy settings TOR_SOCKS_HOST (default 127.0.0.1) and TOR_SOCKS_PORT (default 9050). Connect a client: for Claude Code run claude mcp add darknet-mcp-server -- npx darknet-mcp-server, or with a local clone claude mcp add darknet-mcp-server -- bun run /path/to/darknet-mcp-server/src/index.ts; for Claude Desktop add the npx launch config to mcpServers in claude_desktop_config.json; Cursor / Windsurf use the same JSON format. To use the 7 Tor tools, run a local Tor SOCKS5 proxy first (macOS brew install tor && brew services start tor; Linux sudo apt install tor && sudo systemctl start tor; Docker docker run -d -p 9050:9050 dperson/torproxy) and verify connectivity with npx darknet-mcp-server --check-tor.

.mcp.json
{
  "mcpServers": {
    "darknet": {
      "command": "npx",
      "args": ["-y", "darknet-mcp-server"],
      "env": {
        "HIBP_API_KEY": "optional",
        "INTELX_API_KEY": "optional",
        "ABUSEIPDB_API_KEY": "optional",
        "HUDSONROCK_API_KEY": "optional",
        "HYBRID_API_KEY": "optional",
        "VULNERS_API_KEY": "optional"
      }
    }
  }
}

Shown for Claude Code. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json
{
  "servers": {
    "darknet": {
      "command": "npx",
      "args": [
        "-y",
        "darknet-mcp-server"
      ],
      "env": {
        "HIBP_API_KEY": "optional",
        "INTELX_API_KEY": "optional",
        "ABUSEIPDB_API_KEY": "optional",
        "HUDSONROCK_API_KEY": "optional",
        "HYBRID_API_KEY": "optional",
        "VULNERS_API_KEY": "optional"
      }
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal
claude mcp add darknet -e HIBP_API_KEY=optional -e INTELX_API_KEY=optional -e ABUSEIPDB_API_KEY=optional -e HUDSONROCK_API_KEY=optional -e HYBRID_API_KEY=optional -e VULNERS_API_KEY=optional -- npx -y darknet-mcp-server

Run it in a terminal; replace any <…> placeholders with your own values first.

Check that it works

Run npx darknet-mcp-server --check-tor to verify the Tor proxy, or confirm tools like darknet_list_sources and breachList appear in your client's tool list; npx darknet-mcp-server --list shows all 66 tools.

Troubleshooting

  1. If Tor tools fail, first run `npx darknet-mcp-server --check-tor` to test SOCKS5 connectivity
  2. Verify TOR_SOCKS_HOST / TOR_SOCKS_PORT match the actual Tor listener port (default 127.0.0.1:9050)
  3. A 'set XXX_API_KEY to enable…' style error means that tool requires the key — this is expected graceful degradation
  4. Use `npx darknet-mcp-server --list` to see all tools and `--tool <name> '<json args>'` to debug a single tool
  5. If throttled, configure ABUSECH_AUTH_KEY, OTX_API_KEY, PULSEDIVE_API_KEY and similar to raise quotas
  6. If tools do not appear in your MCP client, confirm the command is `npx darknet-mcp-server` or the local `src/index.ts` path is correct

Things to try

Once connected, you can ask your AI assistant things like:

  • Check if example.com has been in any data breaches and look for stealer logs
  • What ransomware groups are currently targeting the healthcare sector?
  • Search .onion sites for mentions of AcmeCorp and summarize the findings
  • Analyze this file hash for malware indicators: abc123def456 and cross-reference ThreatFox and OTX

Tools 66

tor_status read-only
Check if the local Tor SOCKS5 proxy daemon is running and accessible
tor_fetch_onion read-only
Fetch raw HTML from a .onion URL via Tor SOCKS5 proxy (DNS leak prevention via socks5h)
tor_scrape_onion read-only
Fetch and parse a .onion site — returns structured data: title, links, body text
tor_search_onion read-only
Search for .onion sites using the Ahmia.fi search engine
tor_exit_nodes read-only
Get current Tor exit node IP addresses from the official Tor Project bulk exit list
tor_exit_check read-only
Check if a specific IP address is a known Tor exit node
tor_exit_details read-only
Get detailed Tor exit node information including fingerprints and publish timestamps
ransomwareRecent read-only
Fetch the most recent ransomware victims from ransomware.live
Show 58 more tools
ransomwareGroups read-only
List all known ransomware groups tracked by ransomware.live
ransomwareGroup read-only
Get a detailed profile for a specific ransomware group by name
ransomwareGroupVictims read-only
Get all victims claimed by a specific ransomware group
ransomwareSearch read-only
Search ransomware victims by keyword (company name, domain, etc.)
ransomwareByCountry read-only
Get ransomware victims filtered by ISO 3166-1 alpha-2 country code
ransomwareBySector read-only
Get ransomware victims filtered by sector/industry (healthcare, finance, etc.)
ransomlookGroups read-only
List all 582+ ransomware groups tracked by RansomLook
ransomlookRecent read-only
Fetch the most recent ransomware posts and victim claims from RansomLook
breachList read-only
List all known data breaches from HaveIBeenPwned, optionally filter by domain — free
breachGet read-only
Get details of a specific data breach by name — free
breachLatest read-only
Get the most recently added data breach — free
breachDataClasses read-only
List all data classes (types of compromised data) known to HIBP — free
breachPassword read-only
Check if a password has appeared in known breaches (k-anonymity, only 5-char SHA-1 prefix sent) — free
breachSearch read-only
Search all breaches for a specific account (email/username) — requires HIBP_API_KEY
breachPastes read-only
Search for an email address in publicly posted pastes — requires HIBP_API_KEY
threatfoxGetIocs read-only
Get recent IOCs from ThreatFox reported in the last N days
threatfoxSearch read-only
Search ThreatFox IOCs by IP, domain, hash, or URL
threatfoxTag read-only
Search ThreatFox IOCs by tag (e.g., Cobalt Strike, Emotet)
threatfoxMalware read-only
Search ThreatFox IOCs by malware family using Malpedia naming
urlhausLookup read-only
Look up a URL or host in URLhaus for malware distribution
urlhausTag read-only
Search URLhaus entries by tag
bazaarHash read-only
Look up a malware sample in MalwareBazaar by MD5, SHA1, or SHA256 hash
bazaarRecent read-only
Get the most recently submitted malware samples from MalwareBazaar
bazaarTag read-only
Search MalwareBazaar by tag or YARA signature name
otx_ip read-only
Look up threat intelligence for an IP address — pulse info, reputation, country, ASN
otx_domain read-only
Look up threat intelligence for a domain — pulse info, whois, reputation
otx_hash read-only
Look up threat intelligence for a file hash (MD5, SHA1, SHA256)
otx_cve read-only
Look up threat intelligence for a CVE — related pulses and indicators
otx_search_pulses read-only
Search OTX threat pulses by keyword
abuseipdb_check read-only
Check an IP address for abuse reports — confidence score, ISP, country, report count
abuseipdb_reports read-only
Get individual abuse reports for an IP with detailed comments and categories
abuseipdb_blacklist read-only
Get AbuseIPDB's blacklist of the most reported malicious IP addresses
abuseipdb_check_block read-only
Check an entire CIDR network block for abuse reports
greynoise_ip read-only
Look up an IP on GreyNoise — classification (benign/malicious/unknown), scanner status
greynoise_check read-only
Quick check: is this IP a known scanner or known benign service?
pulsedive_indicator read-only
Look up an indicator (IP, domain, URL, or hash) — risk level, threats, feeds
pulsedive_search read-only
Search Pulsedive indicators by value
pulsedive_explore read-only
Explore linked indicators using advanced queries (related IOCs with risk levels)
stealer_domain read-only
Search stealer log entries by domain — compromised machines, credentials, malware details
stealer_email read-only
Search stealer logs by email address — compromised machines with that email in browser credentials
stealer_ip read-only
Search stealer logs by IP address — compromised machines originating from that IP
vulners_search read-only
Search the Vulners vulnerability database using Lucene queries
vulners_id read-only
Look up a specific vulnerability or exploit by ID (CVE, EDB, GHSA) — free
vulners_exploit read-only
Search specifically for exploits (ExploitDB entries)
btc_address read-only
Look up a Bitcoin address — balance, transaction count, recent transactions
btc_balance read-only
Get Bitcoin address balance in satoshi (quick check without full history)
btc_tx read-only
Get detailed Bitcoin transaction information by hash — inputs, outputs, fees, block info
btc_abuse_check read-only
Check a Bitcoin address for abuse reports on ChainAbuse — scam reports with categories
malware_search read-only
Search Hybrid Analysis sandbox by file hash — verdict, AV detection rate, analysis details
malware_overview read-only
Full malware analysis overview — MITRE ATT&CK techniques, network indicators, processes
malware_feed read-only
Get the latest malware detonation feed — recently analyzed samples with verdicts
onion_lookup read-only
Look up metadata for a .onion address via CIRCL AIL project — first/last seen, status, tags, certs, ports, BTC addresses
intelx_search read-only
Initiate a search on IntelligenceX for leaked data, dark web content, and more
intelx_search_results read-only
Retrieve results for an IntelligenceX search by ID
intelx_phonebook read-only
Phonebook search — find emails, domains, URLs associated with a term
intelx_phonebook_results read-only
Retrieve phonebook search results by ID
phishing_check read-only
Check if a URL is a known phishing site via PhishTank
darknet_list_sources read-only
List all available data sources with configuration status, API key status, and tool counts

Use cases

Assess a domain's or email's exposure across known breaches, pastes and stealer logs
Track ransomware victims and active groups by sector, country or organization
Fetch, scrape and search .onion hidden services through a Tor SOCKS5 proxy
Cross-analyze a file hash for malware: sandbox verdicts, sample repositories, IOCs and threat pulses
Look up abuse reports, scanner status and reputation for IP addresses
Trace Bitcoin address balances, transactions and abuse/scam reports
Search vulnerability databases and ExploitDB entries for exploits

Supported clients

Claude Code
Claude Desktop
Cursor
Windsurf

Listed from the project's documentation, not tested by this site.

Overview

darknet-mcp-server is a Model Context Protocol (MCP) server that gives AI agents dark web and threat intelligence capabilities through 66 tools across 16 data sources. It unifies Have I Been Pwned, IntelligenceX, AlienVault OTX, AbuseIPDB, the abuse.ch suite (ThreatFox, URLhaus, MalwareBazaar), GreyNoise, Pulsedive, Hudson Rock, Vulners, Blockchain.info, ChainAbuse, Hybrid Analysis, CIRCL AIL and the Tor network into a single MCP server, letting an agent query all sources in parallel and correlate data in one conversation for breach investigation, ransomware tracking, dark web reconnaissance, malware analysis and blockchain tracing. It is built with Bun + TypeScript and runs over stdio, starting with zero config via npx.

Similar servers

BioMCP 75 · B

One binary. One grammar. Evidence from the biomedical sources you already trust.

★ 670 · Tools 11 Compare with this →

Source revision 8a19a37a03b3 Data synced 2026-11-02 Read the FMRS scoring method