Firebreak

Connect Firebreak to Claude, Cursor, or any MCP-compatible client. Then just ask:

The AI is the orchestrator.Firebreak doesn't decide what to test — it provides the tools. The AI picks the strategy based on what the user asks and what it finds along the way.

1. User: "Test my app at https://myapp.com" 2. AI calls firebreak_scan_quick({ target_url: "https://myapp.com" }) └── Firebreak probes headers, paths, CORS, TLS └── Returns: 4 findings, score B 3. AI explains results in plain language └── "Found a CORS misconfiguration and 3 missing headers..." 4. User: "Fix the CORS issue" 5. AI calls firebreak_finding_fix({ finding_id: "...", framework: "express" }) └── Returns: before/after code diff for Express 6. User applies fix 7. AI calls firebreak_replay({ finding_id: "..." }) └── Returns: "Fixed. Server no longer reflects arbitrary origins."

These don't hit any external service. They query Firebreak's built-in security knowledge base.

These make HTTP requests to the target. Rate limited and scope-locked.

These work with stored scan results. No external requests.

Every scan runs a subset of these checks depending on the scan type (quick, full, targeted). Each check produces findings with full HTTP evidence.

VCVD — Vibe Coding Vulnerability Database

47 vulnerability patterns that AI-generated code gets wrong. Traditional scanners don't look for these because they're specific to how LLMs write code.

Every scan produces a letter grade based on what was found:

Firebreak is designed for authorized testing only.

Copy.env.exampleto.envand edit as needed.

The release binary is ~8 MB with no runtime dependencies.

Contributions are welcome. Here's how to get involved:

git clone https://github.com/protonese3/Firebreak.git cd Firebreak cargo build cargo test

The server runs onhttp://localhost:9090/mcpby default.

New VCVD patterns— Found a vulnerability pattern that AI consistently generates? Add it tosrc/vcvd/data.rs. Each pattern needs an ID, description, severity, detection hint, and fix.

Scan engine checks— Add new vulnerability checks insrc/engine/checks.rs. Each check function takes a reqwest client, target URL, and safety reference. Return aVec<Finding>.

Best practice guides— Add markdown files toknowledge/best-practices/. Then register them insrc/tools/knowledge/best_practices.rs.

Framework-specific fixes— Thefinding_fixtool generates fix code. Add framework support insrc/tools/analysis.rsin theframework_fix_examplefunction.

Dashboard— The React frontend lives infrontend/. Runnpm run devfor the dev server.
- Fork the repo
- Create a branch (git checkout -b add-new-check)
- Make your changes
- Runcargo check && cargo clippy -- -D warnings
- Open a pull request with a clear description of what you changed and why

- No unnecessary comments. If the code says what it does, don't add a comment.
- Guard clauses over nested if/else.
- Match the style of existing code.
- Every finding must have verifiable evidence — no "this might be vulnerable."

- Web dashboard (React)
- Gray-box scanning with authenticated sessions
- Headless browser for frontend-rendered apps
- Scheduled / recurring scans
- Webhook notifications (Slack, Discord)
- PDF report export
- Cloud-hosted option (managed VPS)
- Plugin system for custom checks
- VCVD community contributions via PR

Do I need to be a security expert to use this?No. That's the point. The AI handles the security expertise. You just tell it what to test.

What MCP clients work with Firebreak?Any client that supports MCP over HTTP: Claude Desktop, Claude Code, Cursor, Windsurf, and others.

Is it safe to run against production?Firebreak is non-destructive (read-only probing), but you should always test against staging first. Rate limiting is on by default.

Can I add my own vulnerability checks?Yes. Add check functions insrc/engine/checks.rsand wire them into the scan methods insrc/engine/mod.rs.

Why Rust?Single binary with no runtime dependencies. Fast. Memory safe. We ship one file and it works.

Why AGPL-3.0?If you run Firebreak as a service for others, you must share your modifications. If you use it internally or self-host for your own team, you don't need to do anything special.

If you build a commercial service on top of Firebreak, the AGPL requires you to open-source your modifications. For internal and self-hosted use, no restrictions beyond the standard AGPL terms.

20 Security Tools · 100 Best Practice Guides · 47 Vulnerability Patterns · 32 Scan Checks

FIREBREAK— Because if you don't test it, someone else will.

Challenge-response quality verification for AI agents and MCP servers.

Cyber Host Artificial Intelligence (C.H.A.I) is Autonomous penetration testing MCP (Model Context Protocol) server with an integrated AI decision engine, multi-provider LLM support, and an extensible plugin architecture.

A Python MCP Server that connects Large Language Models natively to a comprehensive suite of offensive security tools.

A deliberately vulnerable MCP server for hands-on penetration-testing practice — 26 challenges, 78 capture-the-flag flags, plus a victim-agent harness that shows a real LLM agent being exploited.

Detects prompt injection attacks in MCP tool inputs — OWASP LLM Top 10 coverage, real-time scanning, severity scoring for AI agent security

Security scanner for MCP servers — detects prompt injection, credential leaks, and tool poisoning with 52 CVSS-scored rules

pentestMCP: AI-Powered Penetration Testing via MCP, an MCP designed for penetration testers.

An advanced penetration testing tool for automated, LLM-driven security assessments using tools like nmap and dirb.

Paid remote MCP for AI agent safety replay checks, policy gates, eval receipts, control-fix suggestions, and release evidence exports.