IntegrationsMCP Server & Agent Skills

Give your coding agent an autonomous security engine.

Connect BreachLabs to Google Antigravity, Claude, Cursor, Windsurf, Roo/Cline, or any MCP-compatible coding agent. Provide your AI pair-programmer with tools to map attack surfaces, inspect source files, verify reachability with targeted probes, and retest patches automatically.

MCP PROTOCOL v1.0 Agent Skills Ready

Model Context Protocol Streamable HTTP & SSE Live Endpoint: /mcp Core API: /api Deterministic Tooling Zero Telemetry Leakage

RECOMMENDED AUTOMATED NPX INSTALLER

Configure All AI Agents with 1 Command

Run the interactive NPX package installer to automatically detect, merge, and configure BreachLabs MCP tools and security skills into Antigravity, Claude, Cursor, Windsurf, or Cline.

$ npx breachlabs

01 Select Agent Platform

Choose your coding environment

Select your primary AI assistant or IDE to view the exact automated CLI command or manual MCP configuration.

RECOMMENDED

Google Antigravity

Built-in subagent tool registration, sidecars, and autonomous skill triggers.

MCP + Builtin Skill Configure →

Desktop & CLI

Claude (Desktop & Code)

Native claude_desktop_config.json and Claude Code CLI integration.

Stdio Server Configure →

IDE Extension

Cursor

Project-level .cursor/mcp.json and tailored .cursorrules.

MCP Tools + Rules Configure →

Cascade Engine

Windsurf / Cascade

Codeium Cascade agent setup via mcp_config.json and workflow instructions.

Cascade Tools Configure →

VS Code Addon

Roo Code / Cline

Direct MCP tool integration in VS Code via cline_mcp_settings.json.

VS Code Settings Configure →

Any MCP Client

Universal / HTTP & SSE

Direct Streamable HTTP at /mcp, SSE event stream at /mcp/sse, or Docker.

Streamable HTTP / SSE Configure →

STEP 1

Google Antigravity MCP Configuration

Add the BreachLabs Streamable HTTP MCP server to your Antigravity agent configuration (e.g. in ~/.gemini/antigravity/mcp/BreachLabs/).

~/.gemini/antigravity/mcp/BreachLabs/mcp.json

{
  "mcpServers": {
    "breachlabs": {
      "url": "http://127.0.0.1:8000/mcp",
      "transport": "streamable-http"
    }
  }
}

STEP 2

Install Security Skill from skills.sh

Install the BreachLabs skill directly into your agent environment using the universal skills.sh package runner with npx, or copy the universal SKILL.md.

Terminal — Universal Skill CLI (skills.sh)

$npx skills add notaayushsrivastava/BreachLabs
# Installs breachlabs autonomous security engineering skill from skills.sh
# Compatible with Antigravity, Claude, Cursor, Windsurf, Cline & Universal agents

02 Exposed MCP Tools

Autonomous tools, bounded execution, zero open shells.

BreachLabs exposes an allowlisted set of structured tools to your agent. Trigger end-to-end sandbox assessments and receive full evidence-backed security reports directly in MCP.

run_assessment

Executes the complete autonomous assessment pipeline in an isolated sandbox and returns the full Markdown & JSON report.

Inputrepo_path, mode, port

OutputMarkdown + Structured JSON Report

list_routes

Enumerates all HTTP routes, form inputs, JSON endpoints, and authentication flows on the target.

Inputtarget_url

OutputRouteTree JSON

run_static_scan

Performs AST-level taint tracking, secret scanning, and unsafe pattern matching across the repository.

Inputrepo_path

OutputTaintSinks[]

run_dast

Executes scoped runtime HTTP probes against the isolated sandbox service to observe response behavior.

Inputpaths, follow_redirects

OutputHTTPResponse alerts[]

verify_installation

Verifies that both BreachLabs MCP server and AI Skill are properly installed and active in the agent environment.

Inputrepo_path

OutputInstallStatus JSON

get_assessment_report

Retrieves the complete Markdown or structured JSON security assessment report for any assessment ID or latest run.

Inputassessment_id, format

OutputMarkdown / JSON Report

communicate

Bidirectional communication tool for the AI agent to consult the security advisor on triage, exploitability, and remediation diffs.

Inputmessage, topic

OutputAdvisory & Guidance JSON

inspect_repository

Inspects repository structure, entry points, dependency manifests, framework architecture, and returns an AI assessment brief.

Inputpath, repo_path

OutputStructure, Tree & Brief

03 Connection Tester

Test Agent MCP Handshake

Verify JSON-RPC protocol compatibility between your coding agent and the BreachLabs MCP server.

MCP PROTOCOL SIMULATOR READY TO TEST

mcp-jsonrpc-session.log

01[READY] Click 'Test Handshake Simulation' to simulate MCP JSON-RPC protocol negotiation.

Start assessing your applications today.

Build with confidence. Ship software backed by deterministic verification.

Try the Demo View Architecture