Delivery Check: Automated API Testing with AI-Generated Curl Commands

What Is a Delivery Check?

A delivery check is Remocode's automated way to verify that your project actually works. The AI reads your project description, generates 3 to 5 curl commands that test key functionality, runs each command locally, and analyzes the results. You get a clear verdict for each test: PASS, PARTIAL, or FAIL, along with fix suggestions for anything that doesn't work.

Think of it as a smoke test generated by AI — it doesn't replace your test suite, but it catches the "did we actually deploy something that works?" class of issues that are embarrassingly common.

How It Works: Step by Step

Step 1: AI Reads Your Project Description

The delivery check uses your project description (from pane assignment or project configuration) to understand what the application should do. A pane labeled "REST API for user management with CRUD endpoints" tells the AI to test user creation, retrieval, update, and deletion.

Step 2: AI Generates Curl Commands

Based on the project description, the AI generates 3 to 5 curl commands that test the most important functionality. For a user management API, it might generate:

# Test 1: Create a new user
curl -X POST http://localhost:3000/api/users \
  -H "Content-Type: application/json" \
  -d '{"name": "Test User", "email": "test@example.com"}'

# Test 2: Retrieve user list
curl http://localhost:3000/api/users

# Test 3: Get specific user
curl http://localhost:3000/api/users/1

# Test 4: Update user
curl -X PUT http://localhost:3000/api/users/1 \
  -H "Content-Type: application/json" \
  -d '{"name": "Updated User"}'

# Test 5: Delete user
curl -X DELETE http://localhost:3000/api/users/1

The commands are tailored to your specific project, not generic templates. The AI considers the technology stack, common patterns, and the stated functionality.

Step 3: Commands Run Locally

Each curl command is executed on your machine against your locally running application. This tests the actual deployment, not a mock or simulation. If your server isn't running, the tests will fail — which is itself useful information.

Step 4: Results Are Analyzed

The AI examines each curl command's output (HTTP status codes, response bodies, error messages) and assigns a verdict:

• ●PASS — the endpoint responded correctly with expected data
• ●PARTIAL — the endpoint responded but with unexpected data or status codes
• ●FAIL — the endpoint didn't respond, returned an error, or produced clearly wrong results

For PARTIAL and FAIL results, the AI provides fix suggestions explaining what went wrong and how to resolve it.

When to Run Delivery Checks

Before Deployment

Run a delivery check before deploying to staging or production. This catches issues like:

• ●Server not starting due to configuration errors
• ●Endpoints returning 500 errors due to database connection issues
• ●Missing environment variables that the AI agent didn't set up
• ●Routes that were defined but never implemented

After AI Agent Completes a Feature

When an AI agent reports that a feature is complete, run a delivery check to verify. Agents sometimes mark tasks as done when the code compiles, even if the functionality doesn't work end-to-end.

During Code Review

Include delivery check results in your pull request to show reviewers that the feature works. This is more convincing than "all tests pass" because it tests actual HTTP behavior, not unit-level logic.

After Refactoring

Refactoring should preserve existing behavior. A delivery check after a major refactor quickly confirms that the application still responds correctly to the same requests.

Interpreting Results

All PASS

Your application works as described. This doesn't mean it's bug-free, but the core functionality is operational.

Mix of PASS and PARTIAL

Some endpoints work but have minor issues — perhaps wrong status codes (200 instead of 201 for creation) or missing fields in responses. Review the fix suggestions and decide which issues to address.

Any FAIL

Something fundamental isn't working. Common causes:

• ●Server isn't running or is running on a different port
• ●Database isn't connected or migrated
• ●Required environment variables are missing
• ●Routes are defined but handlers throw errors

FAIL results include the curl output and AI analysis, so you can quickly identify the root cause.

Making Delivery Checks More Effective

Write Descriptive Pane Assignments

The quality of generated curl commands depends on the quality of your project description. "Backend API" will produce generic tests. "Express REST API for e-commerce: products CRUD, user auth with JWT, shopping cart with checkout" will produce targeted, meaningful tests.

Ensure Your Server Is Running

Delivery checks test against your locally running application. Make sure the server is started before running the check. If you're running the server in another Remocode pane, check its status first.

Review and Learn from Results

Fix suggestions from failed tests often reveal gaps in your implementation. An AI agent might implement endpoints but forget error handling, input validation, or proper HTTP status codes. Delivery check results highlight these gaps systematically.

Delivery Checks vs. Test Suites

Delivery checks and automated test suites serve different purposes:

| Aspect | Delivery Check | Test Suite | |--------|---------------|------------| | Scope | End-to-end HTTP behavior | Unit and integration logic | | Setup | Zero — AI generates tests | Requires writing test code | | Speed | Seconds for 3-5 tests | Varies with suite size | | Depth | Surface-level smoke testing | Deep logic verification | | Best for | Quick verification | Comprehensive coverage |

Use delivery checks for fast, no-setup verification. Use test suites for thorough coverage. They complement each other — the delivery check catches deployment-level issues that unit tests miss, and unit tests catch logic errors that smoke tests can't detect.