apibench-mcp

A Model Context Protocol (MCP) server for REST and SOAP API testing. Supports HTTP requests, SOAP calls, response assertions, request chaining, data-driven load testing, and response storage.

Overview

apibench-mcp provides a structured set of tools for API testing through MCP-compatible AI clients. It is designed for both functional testing (individual requests with assertions) and performance testing (concurrent load with statistics).

Key capabilities:

• REST requests with authentication, custom headers, and JSONPath extraction
• SOAP requests via WSDL with XPath extraction
• Flag-based response assertions (status code, headers, body content, JSONPath)
• Sequential request chaining with value injection from prior responses
• Data-driven load testing with CSV, JSON, or DSV data files
• System resource preflight checks before load testing
• In-memory response storage with file export

Installation

pip install apibench-mcp

Usage

apibench-mcp

The server requires no arguments. It starts and exposes its tools over the MCP protocol.

MCP Client Configuration

{
  "mcpServers": {
    "api-tester": {
      "command": "apibench-mcp",
      "args": [],
      "autoApprove": [
        "rest_request",
        "soap_request",
        "assert_response",
        "chain_requests",
        "preflight_check",
        "load_test",
        "save_response"
      ]
    }
  }
}

Tools

rest_request

Make an HTTP request (GET, POST, PUT, PATCH, DELETE).

Parameter	Type	Default	Description
method	string	required	HTTP method
url	string	required	Full URL
headers	string or dict	None	HTTP headers
body	string or dict	None	Request body
auth_type	string	None	Authentication type: "bearer", "basic", or "api_key"
auth_value	string	None	Token or key value for authentication
timeout	integer	30	Request timeout in seconds
follow_redirects	boolean	true	Follow HTTP redirects
verify_ssl	boolean	true	Verify SSL certificates
extract	string	None	JSONPath expression to extract from response body

Returns the response with a stored ID for use with assert_response and save_response.

soap_request

Send a SOAP request to a WSDL endpoint.

Parameter	Type	Default	Description
wsdl_url	string	required	URL to the WSDL file
operation	string	required	SOAP operation name
params	string or dict	None	Operation parameters
headers	string or dict	None	Extra HTTP headers
timeout	integer	30	Request timeout in seconds
verify_ssl	boolean	true	Verify SSL certificates
extract	string	None	XPath expression to extract from XML response

assert_response

Validate a stored response against one or more assertions. All assertion parameters are optional; only the ones provided are evaluated.

Parameter	Type	Default	Description
response_id	string	required	ID of a stored response (e.g., "resp_1")
status_code	integer	None	Expected HTTP status code
body_contains	string	None	String that must appear in the response body
body_not_contains	string	None	String that must not appear in the response body
header_exists	string	None	Header name that must be present
header_value	string	None	Expected value for the specified header
jsonpath_expr	string	None	JSONPath expression to evaluate
jsonpath_expected	string	None	Expected value for the JSONPath match

chain_requests

Run a sequence of REST requests where each step can reference extracted values from earlier steps using {{resp_N.extracted[0]}} placeholders.

Parameter	Type	Default	Description
steps	string	required	JSON array of request objects

Each request object supports: method, url, headers, body, auth_type, auth_value, and extract (JSONPath).

preflight_check

Check system resources and recommend a safe maximum concurrency for load testing. Reports CPU count and usage, available memory, file descriptor limits, and a calculated recommendation.

Parameters: None

load_test

Fire N concurrent HTTP requests and return performance statistics.

Parameter	Type	Default	Description
method	string	required	HTTP method
url	string	required	Target URL (supports {{field}} and {{seq}} placeholders)
total_requests	integer	required	Total number of requests to send
concurrency	integer	required	Number of simultaneous requests
headers	string or dict	None	HTTP headers
body	string or dict	None	Static request body
body_template	string	None	Body string with {{field_name}} placeholders
body_template_file	string	None	Path to a file containing the body template
data_file	string	None	Path to CSV, JSON, or DSV file with per-request data
delimiter	string	","	Column delimiter for CSV/DSV data files
timeout	integer	30	Per-request timeout in seconds
verify_ssl	boolean	true	Verify SSL certificates
force	boolean	false	Bypass the safety concurrency cap

Body resolution priority: body_template_file > body_template > body.

When a data_file is provided, each request receives its own row of data. If total_requests exceeds the number of rows, rows are cycled. The {{seq}} placeholder is replaced with a 1-based sequence number per request.

Returns: total requests, success and error counts, error rate, status code distribution, and response time statistics (min, max, avg, median, p95, p99), along with requests per second.

save_response

Save a stored response to a JSON file.

Parameter	Type	Default	Description
response_id	string	required	ID of the stored response
file_path	string	required	Path to save the JSON file

Data-Driven Load Testing

To run load tests with unique data per request:

1. Create a data file (JSON, CSV, or DSV):

[
  {"name": "User-1", "email": "user1@example.com"},
  {"name": "User-2", "email": "user2@example.com"}
]

2. Create a body template:

{"name": "{{name}}", "email": "{{email}}", "role": "tester"}

3. Run the load test referencing both files. Each request will use the next row from the data file, with placeholders replaced by the corresponding field values.

Requirements

• Python 3.12 or later
• Dependencies: mcp[cli], httpx, zeep, jsonpath-ng, lxml, psutil

License

MIT