Bugcrowd MCP Server

A high-performance MCP (Model Context Protocol) server that provides secure, tool-based access to the Bugcrowd API, allowing for natural language interaction through various AI agent platforms.

Features

• Broad API Coverage: Provides tools for interacting with Organizations, Programs, Submissions, Assets, and more.
• Multi-Agent Support: Includes ready-to-use agents for OpenAI, Anthropic (Claude), Google (Gemini), and FastMCP.
• Secure: Uses environment variables for API credentials and performs input validation.
• Dynamic Help: Includes a help() tool that provides real-time documentation for all available tools.

---

Getting Started

1. Installation

# Clone the repository
git clone https://github.com/unstrike/Bugcrowd_MCP_Server.git
cd Bugcrowd_MCP_Server

# Create virtual environment
uv venv
source .venv/bin/activate

# Install all dependencies at once
uv sync

# Or, install packages manually
uv add mcp httpx openai-agents fastmcp anthropic google-generativeai

2. Set Up Credentials

export BUGCROWD_API_USERNAME="your-username"
export BUGCROWD_API_PASSWORD="your-password"

# And the key for your chosen AI platform
export OPENAI_API_KEY="your-key"
export ANTHROPIC_API_KEY="your-key"
export GOOGLE_AI_API_KEY="your-key"

3. Test the Server

Run the included shell script to verify that the MCP server can start and respond to a basic tool call.

./tests/test_server.sh

4. Run the Agent

Run the main agent orchestrator to start an interactive session.

uv run python -m bugcrowd_agents.agent_orchestrator

---

Usage

Interactive Session

Once the agent is running, you can ask it questions or give it commands.

Example Prompts:

• "Show me available bug bounty programs"
• "What are the 5 most recent vulnerability submissions?"
• "Get details for the organization with ID org-123"
• "Use the help tool to see all available commands"

Switching Agents

You can switch between AI platforms by setting the AGENT_PLATFORM environment variable.

# Run with the Gemini agent
AGENT_PLATFORM=gemini uv run python -m bugcrowd_agents.agent_orchestrator

# Run with the Claude agent
AGENT_PLATFORM=claude uv run python -m bugcrowd_agents.agent_orchestrator

# Run with the FastMCP agent
AGENT_PLATFORM=fastmcp uv run python -m bugcrowd_agents.agent_orchestrator

Supported platforms: openai (default), claude, gemini, fastmcp.

Configuring Agent Models (Optional)

You can override the default models for each agent by setting the following environment variables:

• Claude: CLAUDE_MAIN_MODEL, CLAUDE_SUMMARY_MODEL
• Gemini: GEMINI_MAIN_MODEL, GEMINI_SUMMARY_MODEL
• OpenAI: OPENAI_MODEL
• FastMCP: The model is determined by the FASTMCP_PROVIDER and can be configured with the variables above (e.g., CLAUDE_MAIN_MODEL if the provider is anthropic).

Configuring the FastMCP Agent

The fastmcp agent is a powerful, flexible client that can be configured to use different LLM backends (Claude, Gemini, or OpenAI) by setting the FASTMCP_PROVIDER environment variable. This allows you to leverage fastmcp as a versatile intermediary for various AI services.

Supported providers: anthropic (default), google, openai.

Ensure you have the corresponding API key (e.g., GOOGLE_AI_API_KEY) set as an environment variable for the provider you choose.

Examples:

# Run FastMCP with the default Anthropic (Claude) backend
AGENT_PLATFORM=fastmcp uv run python -m bugcrowd_agents.agent_orchestrator

# Run FastMCP with the Google (Gemini) backend
FASTMCP_PROVIDER=google AGENT_PLATFORM=fastmcp uv run python -m bugcrowd_agents.agent_orchestrator

# Run FastMCP with the OpenAI backend
FASTMCP_PROVIDER=openai AGENT_PLATFORM=fastmcp uv run python -m bugcrowd_agents.agent_orchestrator

---

Available Tools

The server provides the following tools. For detailed parameter information, run the agent and type help('<tool_name>').

Category	Tool	Description
Organizations	get_organizations	List all accessible organizations
	get_organization	Get specific organization details
Programs	get_programs	List bug bounty programs
	get_program	Get specific program details
Submissions	get_submissions	List vulnerability submissions
	get_submission	Get specific submission details
	create_submission	Create a new vulnerability report
	update_submission	Update an existing submission
Reports	get_reports	Alternative reports endpoint
	get_report	Get specific report details
Assets	get_customer_assets	List security test targets
	get_customer_asset	Get specific asset details
Rewards	get_monetary_rewards	List bounty rewards
	get_monetary_reward	Get specific reward details
Users	get_users	List users in an organization
	get_user	Get specific user details
Health	server_health	Check server and API connectivity
Help	help	Get detailed help for any tool

---

Advanced Integration

For direct integration with platform-specific CLIs or tools (bypassing the built-in agent handlers), you can use the provided configuration templates.

OpenAI (codex)

The OpenAI codex CLI uses a config.toml file. The docs/config.toml file is a ready-to-use template.

1. Copy docs/config.toml to your ~/.codex/config.toml.
2. In the copied file, update the cwd variable to the absolute path of your Bugcrowd_MCP_Server project directory.
3. Ensure your BUGCROWD_API_USERNAME and BUGCROWD_API_PASSWORD are set as environment variables.

Gemini, Claude, and other MCP-compatible tools

These platforms use a standard config.json file. The docs/config.json file is a ready-to-use template.

1. Copy docs/config.json to the appropriate location for your tool (e.g., ~/.gemini/settings.json).
2. In the copied file, update the cwd variable to the absolute path of your Bugcrowd_MCP_Server project directory.
3. Ensure your BUGCROWD_API_USERNAME and BUGCROWD_API_PASSWORD are set as environment variables.

---

Documentation

• API Reference: A static reference for all tool and endpoint details.
• Architecture Diagram: An overview of the system architecture.
• Bugcrowd REST API: The official API documentation that this server is built upon.

For more detailed information on MCP server configuration, refer to the official documentation for your platform:

• OpenAI: Codex MCP Server Configuration
• Google Gemini: Configure MCP Servers
• Anthropic Claude: MCP for Claude
• FastMCP: JSON Configuration and Running a Server