Decoy Redteam

Autonomous red team for MCP servers. Finds exploitable vulnerabilities before attackers do. Zero dependencies. Zero setup.

Works with: Claude Desktop, Cursor, Windsurf, VS Code, Claude Code, Zed, Cline

🚀 Get Started

npx decoy-redteam            # Dry-run — show attack plan
npx decoy-redteam --live     # Execute attacks against your MCP servers

Decoy Red Team connects to every MCP server on your machine, sends adversarial payloads to their tools, and reports what's exploitable. Not a scanner — an attacker.

🧑‍💻 Install

No install required — run directly with npx. Requires Node.js 18+.

Or pin it in your CI:

- name: Red team MCP servers
  uses: decoy-run/decoy-redteam@v1
  with:
    target: my-server
    token: ${{ secrets.DECOY_TOKEN }}
    sarif: true

🎓 Docs

• Overview
• OWASP Top 10 for Agentic Applications 2026

🗂 What it tests

54 attack patterns across 6 categories:

Every finding maps to OWASP Top 10 for Agentic Applications 2026.

🛡 Safety

Dry-run by default. Running npx decoy-redteam without --live shows what would be tested without executing anything.

Confirmation required. --live prompts for explicit confirmation before executing. No --yes bypass flag.

Safe mode default. Live execution only runs read-only and protocol attacks. Destructive attacks require --live --full with an additional warning.

Browser-automation tools are skipped in safe mode. Tools matching browser_*, navigate, goto, open_url, open_browser, open_tab, open_page, open_window, take_screenshot, screenshot, or screencapture are excluded by default — otherwise SSRF URL payloads cause real browser windows to flicker open for each attack. Opt in with --full.

🛠 Usage

# Dry-run — show attack plan without executing anything
npx decoy-redteam

# Execute attacks against your MCP servers
npx decoy-redteam --live

# Target a specific server
npx decoy-redteam --live --target=my-server

# JSON output for scripting
npx decoy-redteam --live --json

# SARIF output for GitHub Security / CI
npx decoy-redteam --live --sarif

# Only test specific categories
npx decoy-redteam --live --category=input-injection,credential-exposure

Exit codes

🤖 Advanced AI-powered red team (paid plans)

Free decoy-redteam runs 54 deterministic attack patterns. The paid tiers on Decoy Guard (Team $29/user/mo, Business $99/user/mo) add:

• AI-adaptive attacks — LLM-generated payloads specific to your tool schemas
• Encoding bypass suite — 25+ encoding variants per injection vector
• Cross-server chain discovery — finds attack paths across multiple servers
• Exportable HTML reports — branded, print-ready security assessments
• Continuous red teaming — scheduled runs with drift detection

Run with --team --token=YOUR_TOKEN.

📚 Library

import {
  discoverConfigs,
  probeServers,
  planAttacks,
  executeAttacks,
  buildStories,
  calculateCoverage,
  toSarif,
  toJson,
  ATTACKS,
  ENCODINGS,
} from 'decoy-redteam';

🚢 Release Notes

See the hosted changelog.

🤝 Contribute

See CONTRIBUTING.md.

🔗 Related

• decoy-scan — MCP vulnerability scanner (static analysis)
• decoy-tripwire — Tripwire detection for AI agents
• Decoy Guard — Dashboard, monitoring, threat intelligence

📝 License

MIT — see LICENSE.