Automated Delivery Checks in Remocode: AI-Generated Curl Tests for API Validation

# Automated Delivery Checks in Remocode

The delivery check is one of Remocode's most innovative AI commands. It bridges the gap between writing code and verifying it works by having the AI generate tests, execute them, and report results automatically.

How Delivery Checks Work

The delivery check command follows a three-step process:

• ●Test generation. The AI analyzes your recent code changes and terminal activity to understand what was built. Based on this understanding, it generates a set of curl commands designed to test the implemented functionality.

• ●Automated execution. Remocode runs the generated curl commands against your local or specified environment. Each test is executed sequentially, and the responses are captured.

• ●Result reporting. The AI evaluates the responses and classifies each test result:
• ●PASS — The endpoint responded correctly with expected status codes and data.
• ●PARTIAL — The endpoint responded but with incomplete or unexpected results.
• ●FAIL — The endpoint did not respond, returned errors, or behaved incorrectly.

When to Use Delivery Checks

After implementing API endpoints. Once Claude Code, Gemini CLI, or Codex finishes building an API route, run a delivery check to verify the endpoint actually works as intended.

Before marking tasks as done. Delivery checks provide objective verification that functionality is working, not just that code was written.

After refactoring. If you restructure API code, a delivery check quickly validates that existing functionality still works.

During code review. Run a delivery check to demonstrate to reviewers that the implementation has been tested against real requests.

What the AI Tests

The AI is intelligent about what to test. Based on the code context, it typically generates tests for:

• ●Happy path scenarios — Standard requests that should succeed
• ●Required parameters — Verifying that the endpoint handles expected inputs correctly
• ●Response structure — Checking that response bodies contain expected fields
• ●Status codes — Confirming appropriate HTTP status codes (200, 201, 400, 404, etc.)
• ●Authentication — If auth is involved, testing both authenticated and unauthenticated requests

Reading the Results

A typical delivery check output looks like a test report with each test listed alongside its result. For example:

• ●GET /api/users — PASS (200 OK, returned user list)
• ●POST /api/users — PASS (201 Created, user object returned)
• ●GET /api/users/999 — PASS (404 Not Found, appropriate error)
• ●POST /api/users (missing fields) — PARTIAL (400 returned, but error message is generic)
• ●DELETE /api/users/1 — FAIL (endpoint returned 500)

This gives you immediate, actionable feedback. PASS results confirm working functionality. PARTIAL results suggest the feature works but needs refinement. FAIL results indicate bugs that need fixing.

Customizing Delivery Checks

The delivery check prompt can be customized in the Commands tab. Consider adjusting it to:

• ●Specify your local development URL and port
• ●Include authentication tokens or API keys for protected endpoints
• ●Focus on specific endpoint patterns relevant to your current work
• ●Define expected response formats based on your API conventions

Delivery Checks and AI Agents

Delivery checks pair especially well with AI coding agents. A typical workflow looks like this:

• ●Assign a task to Claude Code or Codex (e.g., "build a user registration endpoint")
• ●Let the agent implement the feature
• ●Run status to confirm the agent finished
• ●Run delivery-check to verify the implementation works
• ●Run audit to check for security issues

This creates a complete implement-verify-secure pipeline that catches problems before they reach code review.

Limitations

Delivery checks work best for HTTP APIs where curl commands are a natural testing tool. They are less suited for WebSocket endpoints, GraphQL subscriptions, or purely frontend changes. For those cases, use custom commands with prompts tailored to the appropriate validation method.

The delivery check command turns AI-generated code into AI-verified code, adding a crucial validation step to the AI-assisted development workflow.