Skip to main content
Now that you’ve completed the Getting Started guide and read about Claude Code, this is the recommended reading order. Each phase builds on the last — beginners can follow it straight through, experienced engineers can skip to what they need.

​
Phase 1: Core Concepts

Understand the building blocks:
  1. Workflows - Where you orchestrate the work
  2. Steps - Where the actual work happens
  3. Prompts - How to manage your LLM interactions
  4. Evaluators - Quality control for your AI outputs
  5. API Clients - Call external APIs
Outcome: You understand the architecture

​
Phase 2: Building Real Workflows

Best practices and patterns for building real workflows:
  1. Step Best Practices - Patterns & organization
  2. Evaluator Best Practices - LLM-as-judge patterns
  3. Child Workflows - Compose workflows by calling other workflows
  4. Workflow Context - Understand the execution context
  5. Running Steps in Parallel - How to execute steps concurrently
  6. Credentials - Manage secrets and API keys across environments
Outcome: You can build complex, production-quality workflows

​
Phase 3: Production Readiness

Ship with confidence:
  1. Error Handling - Retries, failures, recovery
  2. Error Hooks - React to failures with custom logic
  3. Testing - Testing deterministic and non-deterministic code
  4. Tracing - Observability and debugging
  5. Cost Estimation - Track and control your AI spend
  6. Evaluation Workflow - Test workflow quality across datasets
Outcome: You’re ready to deploy

​
Phase 4: Integrate with Your App

Connect Output to your product:
  1. Integration Overview - How your app talks to Output over HTTP
  2. API Configuration - Base URL, ports, environment variables
  3. Authentication - Dev vs production auth
  4. Error Responses - Error codes and response handling
Outcome: Your product can trigger and manage workflows via REST

​
Reference (As Needed)