mcp-fuzzer 0.1.2

MCP server fuzzer client and utilities

MCP Fuzzer

A CLI tool for fuzzing MCP server tools using multiple transport protocols, with pretty output using rich.

Features

• Multi-Protocol Support: HTTP, SSE, Stdio, and WebSocket transports
• Tool Discovery: Automatically discovers available tools from MCP servers
• Intelligent Fuzzing: Uses Hypothesis to generate random/edge-case arguments
• Rich Reporting: Beautiful terminal tables with detailed statistics
• Protocol Flexibility: Easy to add new transport protocols

Architecture

The MCP Fuzzer uses a transport abstraction layer to support multiple protocols. Here's how it works:

Installation

pip install mcp-fuzzer

Usage

You can run the fuzzer in several ways:

As a CLI tool (recommended)

mcp-fuzzer --protocol http --endpoint http://localhost:8000/mcp/ --runs 10

As a Python module

python -m mcp_fuzzer --protocol http --endpoint http://localhost:8000/mcp/ --runs 10

As a Python script

python -m mcp_fuzzer.client --protocol http --endpoint http://localhost:8000/mcp/ --runs 10

Supported Protocols

HTTP Transport

mcp-fuzzer --protocol http --endpoint http://localhost:8080/rpc --runs 20

SSE Transport

mcp-fuzzer --protocol sse --endpoint http://localhost:8080/sse --runs 15

Stdio Transport

# Binary executables
mcp-fuzzer --protocol stdio --endpoint "./bin/mcp-shell" --runs 10

# Python scripts
mcp-fuzzer --protocol stdio --endpoint "python3 ./my-mcp-server.py" --runs 10

# Python scripts with spaces in path
mcp-fuzzer --protocol stdio --endpoint '"./My Server/mcp-server.py"' --runs 10

WebSocket Transport

mcp-fuzzer --protocol websocket --endpoint ws://localhost:8080/ws --runs 25

Arguments

• --protocol: Transport protocol to use (http, sse, stdio, websocket)
• --endpoint: Server endpoint (URL for http/sse/websocket, command for stdio)
• --runs: Number of fuzzing runs per tool (default: 10)
• --timeout: Request timeout in seconds (default: 30.0)
• --verbose: Enable verbose logging

Output

Results are shown in a colorized table with detailed statistics:

• Success Rate: Percentage of successful tool calls
• Exception Count: Number of exceptions during fuzzing
• Example Exceptions: Sample error messages for debugging
• Overall Statistics: Summary across all tools and protocols

---

Project dependencies are managed via pyproject.toml.

Test result of fuzz testing of https://github.com/modelcontextprotocol/python-sdk/tree/main/examples/servers/simple-streamablehttp-stateless