@cocaxcode/api-testing-mcp

@cocaxcode/api-testing-mcp

The API testing toolkit for AI coding agents. Test APIs directly from Claude Code, Claude Desktop, Cursor, or any MCP client — using natural language.

"Test my login endpoint with invalid credentials" "Import the API spec and show me all blog endpoints" "Run a load test with 50 concurrent requests to /health" "Generate mock data for the create user endpoint"

Zero cloud dependencies. Everything stored locally as JSON.

Why?

Problem	Solution
Switching between Postman and your IDE	Test APIs without leaving your editor
Remembering endpoint URLs and body schemas	Import your OpenAPI spec — the AI knows your API
Writing {{BASE_URL}}/api/users every time	Just write /api/users — BASE_URL is auto-resolved
Manual regression testing	bulk_test runs your entire collection in seconds
No idea what data to send	mock generates fake data from your API spec
Testing auth flows manually	flow_run chains requests — login → extract token → use it

Quick Start

Install in Claude Code

claude mcp add api-testing -- npx -y @cocaxcode/api-testing-mcp

Install in Claude Desktop / Cursor

Add to your MCP config:

{
  "mcpServers": {
    "api-testing": {
      "command": "npx",
      "args": ["-y", "@cocaxcode/api-testing-mcp"]
    }
  }
}

Setup your environment

"Create an environment called dev with BASE_URL http://localhost:3000 and TOKEN my-dev-token"

That's it. Now just talk:

"GET /api/health"
"Create a blog post with random data"
"Show me all users sorted by creation date"

Features & Examples

🌐 HTTP Requests — request

Execute any HTTP request with full control. Relative URLs (/path) auto-resolve using BASE_URL from the active environment.

"GET /api/users"
"POST /api/blog with title 'Hello World' and content 'My first post'"
"DELETE /api/users/123 with Bearer token abc123"

Supports: GET, POST, PUT, PATCH, DELETE, HEAD, OPTIONS | Headers | Query params | JSON body | Auth (Bearer, API Key, Basic) | Timeout | Variable interpolation {{VAR}}

✅ Assertions — assert

Execute a request AND validate the response in one step. Get structured pass/fail results.

"Assert that GET /api/health returns status 200 and body.status equals 'ok'"
"Verify POST /api/login with wrong password returns 401"
"Check that GET /api/users responds in less than 500ms"

Available operators: eq, neq, gt, gte, lt, lte, contains, not_contains, exists, type

Output:

✅ PASS — 3/3 assertions passed
GET /api/health → 200 OK (42ms)

✅ status === 200
✅ body.status === "ok"
✅ timing.total_ms < 500

🔗 Flow — flow_run

Chain multiple requests in sequence. Extract data from one response and use it in the next step with {{variables}}.

"Run a flow: first login with email [e-mail] and password 123456,
 extract the token from body.token,
 then use that token to GET /api/users"

Output:

✅ FLOW COMPLETO — 2/2 pasos ejecutados

✅ Paso 1: login
   Status: 200 | Tiempo: 145ms
   Extraído: TOKEN=eyJhbGciOiJIUzI1NiIs...

✅ Paso 2: get-users
   Status: 200 | Tiempo: 89ms

📋 OpenAPI Import — api_import, api_endpoints, api_endpoint_detail

Import your Swagger/OpenAPI spec so the AI knows your entire API — endpoints, schemas, required fields, everything.

"Import my API from http://localhost:3000/api-docs-json"
"What endpoints does the blog module have?"
"Show me the details of POST /api/users — what fields does it need?"

Output:

API 'my-backend' imported — 24 endpoints, 15 schemas

Endpoints by tag:
  auth: 3 endpoints
  users: 5 endpoints
  blog: 6 endpoints
  projects: 4 endpoints

🎭 Mock Data — mock

Generate realistic fake data from your OpenAPI spec. Perfect for frontend development without a running backend.

"Generate mock data for creating a user"
"Give me 10 fake blog posts based on the API spec"
"What would the response of GET /api/projects look like?"

Output:

{
  "id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
  "email": "[e-mail]",
  "name": "Test User 73",
  "role": "admin",
  "active": true
}

📊 Load Test — load_test

Launch N concurrent requests and get performance stats: avg, min, max, percentiles, requests/second.

"Run a load test with 50 concurrent requests to GET /api/health"
"Load test POST /api/search with 20 concurrent requests"

Output:

📊 LOAD TEST — GET /api/health

Requests: 50 concurrentes
Exitosos: 50 | Fallidos: 0
Tiempo total: 2145ms
Requests/segundo: 23.31

⏱️ Tiempos de respuesta:
  Min:  45ms
  Avg:  187ms
  p50:  156ms
  p95:  412ms
  p99:  523ms
  Max:  567ms

📋 Status codes:
  200: 50 (100%)

🔄 Diff Responses — diff_responses

Compare two requests side by side. Useful for regression testing or comparing environments (dev vs prod).

"Compare GET /api/users on dev vs prod"
"Diff the response of v1/users vs v2/users"

🚀 Bulk Test — bulk_test

Run all saved requests in your collection and get a summary report. Filter by tag.

"Run all saved requests tagged 'smoke'"
"Bulk test my entire collection"

Output:

✅ BULK TEST — 8/8 passed | 1234ms total

✅ health-check — GET /api/health → 200 (45ms)
✅ list-users — GET /api/users → 200 (123ms)
✅ create-post — POST /api/blog → 201 (89ms)
✅ get-post — GET /api/blog/test-post → 200 (67ms)
...

📤 Export cURL — export_curl

Convert any saved request to a cURL command. Ready to copy-paste and share.

"Export the create-user request as cURL"

Output:

curl \
  -X POST \
  'https://api.example.com/users' \
  -H 'Authorization: Bearer abc123' \
  -H 'Content-Type: application/json' \
  -d '{"name":"John","email":"[e-mail]"}'

💾 Collections — collection_save, collection_list, collection_get, collection_delete

Save, organize, and reuse requests locally. Tag them for easy filtering.

"Save this request as 'create-user' with tags 'users' and 'write'"
"Show me all saved requests tagged 'auth'"
"Delete the old-endpoint request"

🌍 Environments — env_create, env_list, env_set, env_get, env_switch

Manage variables per environment. Switch between dev/staging/prod seamlessly.

"Create a prod environment with BASE_URL https://api.example.com"
"Switch to dev environment"
"Set TOKEN to my-new-token in the current environment"

All 20 Tools

Category	Tool	Description
Request	request	Execute HTTP requests
Assert	assert	Request + validate with assertions
Flow	flow_run	Chain requests, extract variables
Collections	collection_save	Save a request
	collection_list	List saved requests
	collection_get	Get request details
	collection_delete	Delete a request
Environments	env_create	Create environment
	env_list	List environments
	env_set	Set a variable
	env_get	Get variable(s)
	env_switch	Switch active env
API Spec	api_import	Import OpenAPI spec
	api_endpoints	List endpoints
	api_endpoint_detail	Endpoint details
Mock	mock	Generate fake data
Load Test	load_test	Concurrent performance test
Utilities	export_curl	Export as cURL
	diff_responses	Compare responses
	bulk_test	Run entire collection

Storage

All data is stored locally as JSON files in .api-testing/:

.api-testing/
├── active-env
├── collections/
│   ├── health-check.json
│   └── create-user.json
├── environments/
│   ├── dev.json
│   └── prod.json
└── specs/
    └── my-backend.json

Configure the storage directory:

{
  "mcpServers": {
    "api-testing": {
      "command": "npx",
      "args": ["-y", "@cocaxcode/api-testing-mcp"],
      "env": {
        "API_TESTING_DIR": "/path/to/your/.api-testing"
      }
    }
  }
}

You can version these files in git to share with your team.

Development

git clone https://github.com/cocaxcode/api-testing-mcp.git
cd api-testing-mcp
npm install
npm test          # 70 tests
npm run build
npm run typecheck

Test with MCP Inspector

npx @modelcontextprotocol/inspector node dist/index.js

License

MIT — cocaxcode