1. Conduid
  2. Developer Tools
  3. Bbot MCP
MCP server · Developer Tools

Bbot MCP

MCP server for BBOT

Unclaimed MIT last commit a year ago devtools
49Fair

Scored 7 days ago · breakdown

About Bbot MCP

Bbot MCP is an MCP server published by marlinkcyber in the Developer Tools category: mCP server for BBOT.

The repository has 2 stars and 1 forks, with the last commit a year ago. Six months or more without a commit doesn't mean the server is broken, but check the open issues before depending on it in production.

Install

From source
git clone https://github.com/marlinkcyber/bbot-mcp

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Tools (10)

flagsOptional comma-separated list of flags

Parameters not published.

include_progressWhether to include progress updates in the response (default: True)

Parameters not published.

limitMaximum number of results to return (default: 100)

Parameters not published.

modulesOptional comma-separated list of modules to use

Parameters not published.

no_depsDisable dependency installation to prevent sudo prompts (default: True)

Parameters not published.

poll_intervalHow often to check scan status in seconds (default: 5)

Parameters not published.

presetsOptional comma-separated list of presets to apply

Parameters not published.

scan_idThe ID of the scan to wait for

Parameters not published.

targetsComma-separated list of targets (domains, IPs, URLs)

Parameters not published.

timeoutMaximum time to wait in seconds (default: 300 = 5 minutes)

Parameters not published.

Ask AI

✦

Ask AI about Bbot MCP

Powered by Claude · Grounded in docs

I know everything about Bbot MCP. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

Static checks derived from the repository 7 days ago. Heuristics on the source, not an audit.

  • ✓README presentThe repository ships documentation.
  • ✓License declaredUsable under a stated license.
  • ✓Tests presentA test suite exists in the repository.
  • !Dependencies pinnedNo lockfile. Installs can pull different code over time.
  • ✓No dynamic code executionSource scan found no eval, dynamic Function, or shell execution.
  • !Scoped permissionsDoesn't declare a permission scope. Assume it can do anything its process can.

Releases

v0.1.4v0.1.4
v0.1.3v0.1.3
v0.1.1v0.1.1
v0.1.0v0.1.0

README

BBOT MCP Server

A Model Context Protocol (MCP) server for running BBOT security scans. This server provides tools to manage and execute bbot scans through the MCP interface.

Features

  • Module Management: List and explore available bbot modules
  • Preset Management: List and use predefined scan configurations
  • Scan Execution: Start and manage long-running bbot scans
  • Real-time Monitoring: Check scan status and retrieve results
  • Wait & Progress Tracking: Wait for scan completion with timeout and progress reporting
  • Concurrent Scans: Support for multiple simultaneous scans
  • Dependency Management: Comprehensive sudo prevention and no-deps functionality

Installation

From PyPI (Recommended)

pip install bbot-mcp

From Source

git clone https://github.com/marlinkcyber/bbot-mcp.git
cd bbot-mcp
pip install -e .

Using uvx (Run without installing)

uvx bbot-mcp

Install dependencies

It is recommended to install all BBOT dependencies before invoking BBOT MCP server:

bbot --install-all-deps

Usage

Running the MCP Server

After installation, the server can be started using the bbot-mcp command:

bbot-mcp

Or directly with Python:

python -m bbot_mcp.server

Available Tools

The MCP server provides 8 tools for comprehensive bbot scan management:

1. list_bbot_modules()

Lists all available bbot modules categorized by type (scan, output, internal).

2. list_bbot_presets()

Lists all available bbot presets for quick scan configuration.

3. start_bbot_scan(targets, modules="", presets="", flags="", no_deps=True)

Starts a new bbot scan with the specified parameters.

Parameters:

  • targets: Comma-separated list of targets (domains, IPs, URLs)
  • modules: Optional comma-separated list of modules to use
  • presets: Optional comma-separated list of presets to apply
  • flags: Optional comma-separated list of flags
  • no_deps: Disable dependency installation to prevent sudo prompts (default: True)

Example:

start_bbot_scan("example.com,google.com", "httpx,nmap", "web-basic", "safe", True)

Important: The no_deps=True parameter prevents bbot from attempting to install missing dependencies, which would cause sudo password prompts that hang the MCP server.

4. get_scan_status(scan_id)

Retrieves the current status of a specific scan.

5. get_scan_results(scan_id, limit=100)

Retrieves results from a completed or running scan.

Parameters:

  • scan_id: The unique identifier of the scan
  • limit: Maximum number of results to return (default: 100)

6. list_active_scans()

Lists all currently active scans with their basic information.

7. wait_for_scan_completion(scan_id, timeout=300, poll_interval=5, include_progress=True)

Waits for a scan to complete with timeout and progress reporting.

Parameters:

  • scan_id: The ID of the scan to wait for
  • timeout: Maximum time to wait in seconds (default: 300 = 5 minutes)
  • poll_interval: How often to check scan status in seconds (default: 5)
  • include_progress: Whether to include progress updates in the response (default: True)

Returns:

  • Success response with completion details, elapsed time, and progress updates
  • Timeout response if scan doesn't complete within the specified time
  • Error response for invalid scan IDs or other issues

Example:

# Wait for scan to complete with custom timeout
result = wait_for_scan_completion("scan-123", timeout=600, poll_interval=10)

8. get_dependency_info()

Provides information about bbot's dependency management system and how the MCP server handles dependencies.

MCP config example:

{
  "mcpServers": {
    "bbot": {
      "command": "uvx",
      "args": ["--refresh","bbot-mcp"]
    }
  }
}

Scan Management

Scan Lifecycle

  1. Starting: Scan is being initialized
  2. Running: Scan is actively executing
  3. Completed: Scan finished successfully
  4. Error: Scan encountered an error

Long-running Scans

Scans run in separate threads to avoid blocking the MCP server. You can:

  • Start multiple scans concurrently
  • Check status while scans are running
  • Retrieve partial results from ongoing scans

Development

Testing

Run the test suite to verify functionality:

# Run all tests
pytest

# Run specific test categories
python tests/test_server.py
python tests/simple_test.py
python tests/test_imports.py

# Test the binary command
bbot-mcp --help

Project Structure

bbot-mcp/
├── bbot_mcp/              # Main package
│   ├── __init__.py        # Package initialization
│   └── server.py          # MCP server implementation
├── tests/                 # Test suite
├── pyproject.toml         # Package configuration
├── README.md             # This file
└── requirements.txt      # Development dependencies

Example MCP Client Usage

# Connect to the MCP server and use the tools
client = MCPClient("bbot-scanner")

# List available modules
modules = client.call_tool("list_bbot_modules")

# Start a scan
scan_result = client.call_tool("start_bbot_scan", {
    "targets": "example.com",
    "presets": "web-basic"
})

# Check scan status
status = client.call_tool("get_scan_status", {
    "scan_id": scan_result["scan_id"]
})

# Wait for scan to complete
completion = client.call_tool("wait_for_scan_completion", {
    "scan_id": scan_result["scan_id"],
    "timeout": 300
})

# Get results when complete
results = client.call_tool("get_scan_results", {
    "scan_id": scan_result["scan_id"],
    "limit": 50
})

Security Considerations

  • This tool is designed for authorized security testing only
  • Always ensure you have permission to scan target systems
  • Be aware that bbot scans can be resource-intensive and may take significant time
  • Some modules may be considered intrusive - use the --allow-deadly equivalent flags carefully

Dependency Management

The MCP server includes comprehensive dependency management to prevent sudo password prompts:

Automatic Protection Measures

  • Default Behavior: no_deps=True - Dependencies are disabled by default
  • Environment Variables: Multiple layers of sudo prevention (SUDO_ASKPASS, DEBIAN_FRONTEND, etc.)
  • Stdin Redirection: Blocks all interactive input to prevent hanging
  • Module Exclusions: Problematic modules (sslcert, trufflehog) are automatically excluded
  • Force Configuration: Modules run even if dependencies fail

Key Features

  • Comprehensive Sudo Prevention: Multiple environment variables and configurations prevent any sudo prompts
  • Graceful Degradation: Scans continue even when some modules can't load dependencies
  • Pre-installation Support: Install dependencies manually if needed: pip install <module-deps>
  • macOS Compatibility: Special handling for Homebrew vs APT package manager conflicts

Excluded Modules

You can exclude problematic modules by setting following environment variable:

export BBOT_EXCLUDE_MODULES="trufflehog,sslcert"

Example exclude mentioned modules if you have any dependency issues (i.e. on Mac OS X):

  • sslcert: APT dependency incompatible with macOS Homebrew
  • trufflehog: Dependency installation conflicts

Override Option: Set no_deps=False only if you're certain no sudo prompts will occur

For more information about bbot itself, visit: https://github.com/blacklanternsecurity/bbot

README fetched from the repository 7 days ago. The original is authoritative.

Questions

About Bbot MCP

How do I install Bbot MCP?

No package has been published for this server, so install it from source: clone the repository and follow its README.

Is Bbot MCP safe to use with an AI agent?

Its trust score is 49 out of 100 (fair). It passes 4 of the 6 static security checks run so far; the failures are listed above. It has no ConduID identity yet, so agent calls to it are not receipted.

Is Bbot MCP still maintained?

The last commit was a year ago. That's long enough that you should check whether the maintainer is responding to issues before depending on it.