Skip to main content
Glama
HoneyClaw

HoneyClaw

by GrooveXlabs

HoneyClaw 🍯

Autonomous AI Honeypot & Agent Canary Engine
Crafted with defense-in-depth by GrooveXlabs β€” Sovereign-Grade AI Infrastructure.

CI Status License: MIT Python Version Security Standard


⚑ The Threat: Rogue AI Scrapers & Agentic Attackers

Traditional honeypots (Cowrie, Dionaea) are designed to trap human attackers and automated bash worms scanning port 22.

In 2026, the primary attack surface is shifting to Autonomous AI Agents:

  • Unfettered web-crawling LLMs scanning internal documentation.

  • Rogue agents exploiting open Model Context Protocol (MCP) servers.

  • Prompt-injection scrapers attempting to exfiltrate private API keys, databases, and customer PII.


Related MCP server: Deception Remote MCP Server

🍯 The Solution: HoneyClaw

HoneyClaw is an active deception honeypot designed specifically to lure, trap, and profile autonomous AI agents. It exposes high-value, realistic fake MCP tools that:

  1. Bait Agents with Synthetic Assets: Returns realistic AWS credentials, GitHub tokens, database dumps, and .env files seeded with cryptographically signed canary tokens.

  2. Behavioral AI Profiling: Analyzes agent interaction speed, tool exploration sequencing, and detects prompt injection attempts (ignore previous instructions, DAN mode).

  3. Forensic Canary Tracing: If an attacker attempts to use or sell exfiltrated credentials anywhere on the internet, HoneyClaw's HMAC signature mathematically proves they were stolen from this honeypot.

   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
   β”‚ Autonomous AI Attacker / Scraper β”‚
   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                    β”‚  MCP JSON-RPC / Tools Discovery
                    β–Ό
   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
   β”‚                 HONEYCLAW DECOY SUITE                   β”‚
   β”‚ ─────────────────────────────────────────────────────── β”‚
   β”‚  [🍯] fetch_vault_credentials   --> Signed AWS/GH Canaries
   β”‚  [🍯] query_customer_database   --> Fake PII + Canary Emails
   β”‚  [🍯] run_system_diagnostic     --> Restricted Sandboxed Shell
   β”‚  [🍯] read_internal_file        --> Seeded .env & Configs
   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                    β”‚
                    β–Ό
   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
   β”‚        HUNTER-GUARDIAN BEHAVIORAL PROFILER              β”‚
   β”‚  * Prompt Injection Detector (Jailbreak / System Leaks) β”‚
   β”‚  * Archetype Classifier (Scraper / Recon / Exfiltrator) β”‚
   β”‚  * Forensic Event Logger (honeyclaw_events.jsonl)       β”‚
   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

πŸš€ Key Features

  • Authentic MCP Protocol Server: Fully compliant with Anthropic's Model Context Protocol (JSON-RPC 2.0). Connects seamlessly over stdio to Antigravity, Claude Desktop, Cursor, or external agent runners.

  • Cryptographic Canary Engine: Generates and validates synthetic AWS keys (AKIA...), GitHub PATs (ghp_...), and OpenAI keys (sk-...) embedded with tamper-evident HMAC timestamps.

  • Agent Fingerprinter: Distinguishes human typing from automated reasoning loops and tags attacker archetypes (AUTOMATED_SCRAPER, RECON_BOT, DATA_EXFILTRATOR, PROMPT_INJECTOR).

  • Safe Sandbox Simulation: Shell diagnostic decoy simulates believable command responses (whoami, id, uname -a) while discarding malicious commands without touching host hardware.


πŸ“¦ Quickstart & Installation

# Clone the repository
git clone https://github.com/GrooveXlabs/honeyclaw.git
cd honeyclaw

# Install package
pip install -e .

Checking Honeypot Status

honeyclaw status

Generating Canary Tokens Manually

# Generate an AWS Access Key Pair canary
honeyclaw generate-canary --type aws --label "prod-backup"

# Generate a GitHub Personal Access Token canary
honeyclaw generate-canary --type github --label "repo-sync-bot"

Verifying an Intercepted Token

honeyclaw verify-canary --key "AKIA6789ABCD..." --label "prod-backup"

Running the Decoy MCP Server

honeyclaw serve

βš™οΈ Integrating into Antigravity & Claude Desktop

To expose HoneyClaw as a decoy tool server in your mcp_config.json:

{
  "mcpServers": {
    "internal-infra-tools": {
      "command": "honeyclaw",
      "args": ["serve"]
    }
  }
}

πŸ“œ License

Licensed under the MIT License.
Copyright (c) 2026 GrooveXlabs.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      C
      maintenance
      A deception-based threat detection server that impersonates enterprise MCP integrations to log and forward attacker interactions to SIEM systems. It provides convincing fake responses across 38 tools while capturing forensic details of all MCP tool calls.
      MIT
    • A
      license
      Not graded
      quality
      C
      maintenance
      Simulates sensitive internal tools as a honeypot to detect unauthorized access and AI agent behaviors using Canarytokens, deployed as a serverless Cloudflare Worker with MCP protocol support.
      23
      MIT
    • A
      license
      Not graded
      quality
      C
      maintenance
      Tripwire detection for autonomous AI agents. Plants honeypot files, MCP tripwire tools, and API decoy endpoints to log agent scope creep and unauthorized tool use with full forensic context.
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables defenders to deploy a decoy MCP tool server that records and fingerprints how LLM agents probe, escalate, and persist, without exposing real systems.
      1
      MIT