API Gateway
Last reviewed: August 2026
Overview
Section titled “Overview”When exposing backend services externally as APIs, implementing authentication, rate limiting, request transformation, and monitoring separately for every service creates duplication. API Gateway acts as a front door that handles these common concerns in one place.
This is similar to putting authentication/routing logic into a reverse proxy (Nginx, HAProxy) on-premises, but a cloud API Gateway is managed and provides integrated scaling, monitoring, and even a developer portal.
Key Features
Section titled “Key Features”- Authentication/authorization — Access control via API keys, OAuth, JWT, IAM
- Rate limiting (throttling) — Protects the backend by limiting requests per second
- Request/response transformation — Adding headers, body mapping, protocol conversion
- Caching — Caches responses for repeated requests
- Monitoring — Automatically collects request counts, latency, error rates
- Developer portal — Auto-generated API documentation, key issuance
Authentication/Authorization Integration
Section titled “Authentication/Authorization Integration”API Gateway itself doesn’t handle authentication logic — it integrates with an external authentication service.
| Vendor | Authentication service | Notes |
|---|---|---|
| AWS | Cognito | User pools + social login. Native integration with API Gateway |
| AWS | Lambda Authorizer | Implement custom authentication logic in Lambda |
| Azure | Entra ID (formerly Azure AD) | OAuth 2.0 / OpenID Connect |
| Azure | APIM Policy (validate-jwt) | Configure JWT token validation as a policy |
| Google Cloud | Firebase Auth / Identity Platform | Social login, multi-factor authentication |
| Google Cloud | Service Account + IAM | Service-to-service authentication |
| OCI | OCI IAM / Identity Domains | OAuth 2.0, JWT validation |
Additional Features
Section titled “Additional Features”| Feature | AWS | Azure | Google Cloud | OCI |
|---|---|---|---|---|
| Usage plans/quotas | Usage Plans + API Keys | Subscription + Quota Policy | Apigee Rate Limiting | Rate Limiting Policy |
| Request validation | Request Validator (model schema) | APIM Policy (validate-content) | Apigee OAS Validation | Request Validation Policy |
| Custom domain | Custom Domain + ACM certificate | Custom Domain + Managed Certificate | Custom Domain + SSL | Custom Domain + SSL certificate |
| WebSocket | WebSocket API | — (separate SignalR) | — (separate Firebase Realtime) | — |
| GraphQL | AppSync | — (3rd party) | — (3rd party) | — |
Product Comparison
Section titled “Product Comparison”| Vendor | Product | Notes |
|---|---|---|
| AWS | API Gateway (REST/HTTP/WebSocket) | Native integration with Lambda. Ideal for serverless API composition |
| AWS | AppSync | GraphQL-only. Supports real-time subscriptions |
| Azure | API Management (APIM) | Built-in developer portal. Unifies multicloud/hybrid APIs. AI Gateway tier (Public Preview, Jun 2026) for AI model and MCP server governance |
| Google Cloud | Apigee | Enterprise API management platform. Analytics/monetization features |
| Google Cloud | API Gateway | Lightweight. Good fit for Cloud Functions/Cloud Run integration |
| OCI | OCI API Gateway | OCI Functions integration. Supports authentication, rate limiting, request transformation |
Key Differences
Section titled “Key Differences”AWS API Gateway — Has the deepest integration with Lambda, optimized for exposing serverless backends as APIs. It’s split into 3 types for different purposes (REST API / HTTP API / WebSocket API), and HTTP API costs about 1/3 of REST API. Most new projects start with HTTP API, choosing REST API only when caching, WAF, or Usage Plans are needed.
Azure API Management — A full-featured platform with a built-in developer portal, API versioning, and policy engine. A single service handles REST, WebSocket, and GraphQL. On-premises and cloud APIs can be unified under a single gateway. The AI Gateway tier (Public Preview, Jun 2026) provides purpose-built publishing, security, and governance for AI models and MCP servers. All v2 tiers (Basic v2, Standard v2, Premium v2) are now GA with flexible pricing and performance options.
Google Cloud Apigee — An enterprise platform that manages APIs as a product. A single service handles all protocols, with strengths in API usage analytics, monetization, and partner management. Gemini Code Assist integration for natural-language API spec generation is now GA, and API Hub supports agent registration, assessment, and Git-based synchronization.
OCI API Gateway — Natively integrates with OCI Functions, and a single service lets you configure authentication (JWT validation), rate limiting, and request transformation as policies.
API Deployment Stages and Version Management
Section titled “API Deployment Stages and Version Management”Since an API becomes hard to change once deployed to production, staged deployment (dev→staging→production) and version management are important.
| Feature | AWS API Gateway | Azure APIM | Google Cloud Apigee | OCI API Gateway |
|---|---|---|---|---|
| Stage/environment | Stages (dev, staging, prod) | Environments | Environments (test, prod) | Deployments |
| Canary deployment | Canary Deployment (weight-based) | Revision + Release | Revision + TargetServer | Route Rule (weight-based) |
| Version management | API Version (v1, v2 as separate endpoints) | API Revision + Version | API Revision + Version | Spec Version |
| Rollback | Switch to a previous Deployment | Switch Revision | Switch Revision | Switch to a previous Deployment |
Progressive Deployment Strategies
Section titled “Progressive Deployment Strategies”| Strategy | Description | Vendor implementation |
|---|---|---|
| Canary | Send a small amount of traffic (5-10%) to the new version, monitor, then expand | AWS: API Gateway Canary Deployment (Stage weighting), Azure APIM: Revision + Traffic Split, Google Cloud Apigee: TargetServer weighting |
| Blue/Green | Run two environments simultaneously, then switch traffic. Immediate rollback possible | AWS: switch Stage, Azure APIM: switch Revision, Google Cloud: switch Revision |
| Version separation | Coexist as separate /v1, /v2 endpoints |
Supported by every vendor. New version added without affecting existing consumers |
OpenAPI (Swagger) Integration
Section titled “OpenAPI (Swagger) Integration”Teams that defined API specs with Swagger on-premises can maintain the same workflow in the cloud.
- Import: Automatically generate the API Gateway from an OpenAPI spec (supported by AWS, Azure APIM, and Apigee)
- Export: Extract the OpenAPI spec from the API Gateway for documentation
- IaC integration: Define the OpenAPI spec inline in CloudFormation/Terraform to manage the API configuration as code
- Developer portal: Azure APIM and Apigee automatically generate interactive API documentation from the OpenAPI spec
API Testing Tools
Section titled “API Testing Tools”| Tool | Role | Relationship with API Gateway |
|---|---|---|
| Postman | Manual API testing, collection management, environment variables | Manage per-stage URLs as environments, auto-refresh auth tokens |
| Swagger UI | Interactive documentation based on the OpenAPI spec | Auto-generated from the spec exported by API Gateway |
| Vendor console testing | AWS console Test tab, APIM Test tab | Quick validation before deployment |
| curl / httpie | Fast CLI-based testing | Smoke testing in a CI/CD pipeline |
Cloud APIs have complex authentication flows (OAuth, API Key, IAM Sig v4) and frequently switch between environments (dev/staging/prod), so it’s efficient to manage environment variables and authentication systematically with a tool like Postman.
Common Mistakes
Section titled “Common Mistakes”- Starting with REST API without distinguishing it from HTTP API — On AWS, HTTP API costs 1/3 and meets most requirements. Choose REST API only when caching, WAF, or Usage Plans are needed.
- Changing an existing response format without API versioning — Existing consumers (clients) break immediately. Add a new version (
/v2) and keep the existing version running. - Mistaking API Gateway’s Canary for a backend code deployment canary — API Gateway Canary is about gateway configuration changes. Backend code deployment needs to be configured separately, via Lambda Alias weighting or an LB Target Group.
Checklist
Section titled “Checklist”- Is authentication/authorization (API Key, JWT, IAM) applied to every endpoint?
- Is rate limiting (throttling) configured to prevent backend overload?
- Is the OpenAPI spec synced with IaC so the API configuration is managed as code?
References
Section titled “References”- Amazon API Gateway Documentation
- API Gateway Authentication/Authorization
- Amazon Cognito Documentation
- AWS AppSync Documentation