Skip to main content
Start with one service and a request whose result you can inspect. Once authentication and read access work, follow that service’s guide to create and update resources. For a complete local starting point, use Your First Operation. It creates a known order and stock fixture, verifies it from a second process, and generates the configuration and task prompt for an AI agent to inspect those same records.

Choose your interface

Each service has its own API contract. Use the endpoint’s request body and authentication header; the embedded library’s methods and objects are a separate interface.

Make a first request from your application

The examples below read the ResponseCX workspace. They use standard HTTP libraries and need a ResponseCX API key with agents:read. Create a key in Settings → API keys and export it in the terminal that will run your application:
Use Node.js 18 or later. Save this as workspace.mjs and run node workspace.mjs.
Success: the response includes agent_count, an agents array, and next_action. Counts and suggestions depend on your workspace. See the workspace contract for the complete response, then follow the ResponseCX quickstart to create an agent.

Build your first write

  1. Read the endpoint’s Request body, Access, and Status codes sections.
  2. Create one resource and save its returned ID. Use that ID in later requests.
  3. Read it back and verify the fields you changed.
  4. Where the endpoint supports idempotency, save one key per intended operation and reuse it with the same payload after a timeout. Changing the payload requires a new operation and key.
  5. Follow the service’s activation or deployment steps after configuration is complete.
For ResponseCX, create an agent supports an Idempotency-Key; update an agent has a different contract. Check support per endpoint before building a general retry wrapper.

Handle failures deliberately

Error bodies vary by service. The cross-engine error guide shows which field to parse; the support checklist lists the information needed to investigate a failing request. The error-handling guide adds a runnable example and explains how to distinguish a retryable failure from an unresolved write or business rejection.

Follow a complete integration

For local commerce, start with the customer workflow, then run the integration test suite to verify stock and retry behavior in isolated databases.

Customer support agent

Create and inspect a ResponseCX agent, then add knowledge, rules and tools.

System integration

Connect Shopify and NetSuite and inspect a sync run.

Durable automation

Configure, start and inspect a workflow through the Temporal Engine.

Policy-controlled action

Evaluate a refund policy and verify the result before executing the action.
Last modified on September 20, 2026