StackHawk MCP Server

Current Version: 1.2.5 Requires Python 3.10 or higher

A Model Context Protocol (MCP) server for integrating with StackHawk's security scanning platform. Helps developers set up StackHawk, run security scans, and triage findings to fix vulnerabilities — all from within an LLM-powered IDE or chat.

Table of Contents

• Features
• Installation
• Usage
• Configuration
• Available Tools
• Testing & Development
• Example Configurations
• Integrating with LLMs and IDEs
• Contributing
• License

Features

• Setup: Detect your project, create a StackHawk application, and generate a ready-to-scan stackhawk.yml
• Scan: Run StackHawk scans directly from your IDE or chat (with install help if the CLI is missing)
• Triage: Get actionable findings at or above your failure threshold for remediation
• Validate: Check YAML configs against the official schema and validate field paths to prevent hallucination
• Custom User-Agent: All API calls include a versioned User-Agent header

Installation

1. Install via pip (make sure you have write permission to your current python environment):
   bash

   Copy

   > pip install stackhawk-mcp
   # Requires Python 3.10 or higher
   

Or Install via pip in a virtual env:

> python3 -m venv ~/.virtualenvs/mcp
> source ~/.virtualenvs/mcp/bin/activate
> (mcp) pip install stackhawk-mcp
# Requires Python 3.10 or higher

Or Install via pip using pyenv:

> pyenv shell 3.10.11
> pip install stackhawk-mcp
# Requires Python 3.10 or higher

Or Install locally from this repo:

> pip install --user .
# Run this command from the root of the cloned repository

2. Set your StackHawk API key:
   server.ts

   Copy

   > export STACKHAWK_API_KEY="your-api-key-here"
   

Usage

Running the MCP Server

python -m stackhawk_mcp.server

Running the HTTP Server (FastAPI)

python -m stackhawk_mcp.http_server

Running Tests

pytest

Integrating with LLMs and IDEs

StackHawk MCP can be used as a tool provider for AI coding assistants and LLM-powered developer environments, enabling security scanning setup, YAML validation, and vulnerability triage directly in your workflow.

Cursor (AI Coding Editor)

• Setup:
  • Follow the installation instructions above to install stackhawk-mcp in your python environment.
  • In Cursor, go to Cursor Settings->Tools & Integrations->MCP Tools
  • Add a "New MCP Server" with the following json, depending on your setup:
    • Using a virtual env at ~/.virtualenvs/mcp:
      config.json

      Copy

      {
        "mcpServers": {
          "stackhawk": {
            "command": "/home/bobby/.virtualenvs/mcp/bin/python",
            "args": ["-m", "stackhawk_mcp.server"],
            "env": {
              "STACKHAWK_API_KEY": "${env:STACKHAWK_API_KEY}"
            },
            "disabled": false
          }
        }
      }
      

    • Using pyenv:
      config.json

      Copy

      {
        "mcpServers": {
          "stackhawk": {
            "command": "/home/bobby/.pyenv/versions/3.10.11/bin/python3",
            "args": ["-m", "stackhawk_mcp.server"],
            "env": {
              "STACKHAWK_API_KEY": "${env:STACKHAWK_API_KEY}"
            },
            "disabled": false
          }
        }
      }
      

    • Or use python directly:
      config.json

      Copy

      {
        "mcpServers": {
          "stackhawk": {
            "command": "python3",
            "args": ["-m", "stackhawk_mcp.server"],
            "env": {
              "STACKHAWK_API_KEY": "${env:STACKHAWK_API_KEY}"
            }
          }
        }
      }
      

    • Then make sure the "stackhawk" MCP Tool is enabled
• Usage:
  • Use Cursor's tool invocation to call StackHawk MCP tools (e.g., vulnerability search, YAML validation).
  • Example prompt: Validate this StackHawk YAML config for errors.

OpenAI, Anthropic, and Other LLMs

• Setup:
  • Deploy the MCP HTTP server and expose it to your LLM system (local or cloud).
  • Use the LLM's tool-calling or function-calling API to connect to the MCP endpoint.
  • Pass the required arguments (e.g., org_id, yaml_content) as specified in the tool schemas.
• Example API Call:
  config.json

  Copy

  {
    "method": "tools/call",
    "params": {
      "name": "validate_stackhawk_config",
      "arguments": {"yaml_content": "..."}
    }
  }
  

• Best Practices:
  • Use anti-hallucination tools to validate field names and schema compliance.
  • Always check the tool's output for warnings or suggestions.

IDEs like Windsurf

• Setup:
  • Add StackHawk MCP as a tool provider or extension in your IDE, pointing to the local or remote MCP server endpoint.
  • Configure environment variables as needed.
• Usage:
  • Invoke setup, scanning, validation, and triage tools directly from the IDE's command palette or tool integration panel.

General Tips

• Ensure the MCP server is running and accessible from your LLM or IDE environment.
• Review the Available Tools & API section for supported operations.
• For advanced integration, see the example tool usage in this README or explore the codebase for custom workflows.

GitHub Copilot Agents

StackHawk can be added to the GitHub Coding Agent as an MCP server or as its own GitHub Custom Agent.

Add to GitHub Coding Agent

You can add StackHawk MCP to the GitHub Copilot Coding Agent. This gives the agent all the stackhawk/ tools.

StackHawk MCP installation into the Coding Agent

General instructions on GitHub

For StackHawk MCP, the MCP Configuration JSON should look something like this:

{
  "mcpServers": {
    "stackhawk": {
      "type": "local",
      "tools": [
        "*"
      ],
      "command": "uvx",
      "args": [
        "stackhawk-mcp"
      ],
      "env": {
        "STACKHAWK_API_KEY": "COPILOT_MCP_STACKHAWK_API_KEY"
      }
    }
  }
}

Then in the Repository's Settings->Environments->copilot->Environment Secrets, add COPILOT_MCP_STACKHAWK_API_KEY with your StackHawk API Key.

Installation verification instructions

StackHawk Onboarding Agent as a GitHub Copilot Custom Agent

You can the StackHawk Onboarding Agent as a custom agent at the enterprise, organization, or repository level in GitHub. When added, the StackHawk Onboarding Agent becomes a selectable option in the Copilot Agent Chat with context to help with onboarding, plus it installs stackhawk-mcp so the agent has access to all of those tools.

StackHawk Onboarding Agent installation

The general approach is to take the StackHawk Onboarding Agent defintion and apply it to either the desired repository, enterprise, or organization in GitHub.

• Instructions for installing into a repository on GitHub
• Instructions for installing into an enterprise on GitHub
• Instructions for installing into an organization GitHub

Note that the mcp-servers block in the StackHawk Onboarding Agent definition references an environment variable called COPILOT_MCP_STACKHAWK_API_KEY. Go to the Repository's Settings->Environments->copilot->Environment Secrets, add COPILOT_MCP_STACKHAWK_API_KEY with your StackHawk API Key.

Configuration

• All HTTP requests include a custom User-Agent header:
  Code

  Copy

  User-Agent: StackHawk-MCP/{version}
  

• The version is set in stackhawk_mcp/server.py as STACKHAWK_MCP_VERSION.
• Set your API key via the STACKHAWK_API_KEY environment variable.

Available Tools

The MCP server exposes 7 tools organized around the developer workflow:

Example Tool Usage

# Set up StackHawk for a project
result = await server.call_tool("setup_stackhawk_for_project", {"host": "http://localhost:3000"})

# Validate a YAML config
result = await server.call_tool("validate_stackhawk_config", {"yaml_content": "..."})

# Run a scan
result = await server.call_tool("run_stackhawk_scan", {})

# Get findings to triage
result = await server.call_tool("get_app_findings_for_triage", {})