api-design

name	api-design
description	API design, REST, GraphQL, or authentication patterns.

API Design Patterns

Language-agnostic patterns for designing REST and GraphQL APIs. Focus on solving real problems with simple, maintainable solutions that match the project's existing API conventions.

Reference Documentation

For detailed patterns and examples, see:

• rest-patterns.md - Resource naming, HTTP methods, status codes, versioning, pagination
• graphql-patterns.md - Schema design, queries, mutations, resolvers, N+1 prevention
• auth-patterns.md - API keys, JWT, OAuth2, permission models (RBAC/ABAC)

---

Philosophy

• Simple over clever - Choose straightforward patterns that solve the problem
• Consistency - Apply patterns consistently across endpoints
• Least Astonishment - APIs should do what their names suggest, nothing more
• Pragmatism - Pick approaches based on actual use cases, not theoretical purity
• No over-engineering - Don't add features or complexity "just in case"

---

Avoid complexity theater

Adding patterns is easy. Adding patterns WORTH THE COMPLEXITY is hard.

Before recommending ANY API pattern, ask:

1. Does this solve a real problem or a hypothetical one?

   • "We might need..." vs "We currently need..."
   • YAGNI (You Aren't Gonna Need It) applies to APIs too

2. Is the simpler approach sufficient?

   • REST before GraphQL (unless you have N+1 query problems)
   • Query parameters before complex filtering DSLs
   • Flat responses before nested structures

3. What's the maintenance cost?

   • Every abstraction requires documentation, testing, and support
   • Versioning strategies add cognitive overhead

The Complexity Theater Litmus Test

"If I remove this pattern, what specific problem occurs in production?"

If the answer is vague ("flexibility", "future-proofing", "best practices"), the pattern may be theater.

API Anti-Patterns to Avoid

Anti-Pattern	Example	Problem
Premature GraphQL	"Use GraphQL for flexibility"	When you have 3 endpoints and 10 users
Over-versioning	"v1, v2, v3 for every change"	When backwards-compatible changes suffice
Enterprise patterns	"Add HATEOAS for discoverability"	When your API has 5 endpoints
Pagination theater	"Cursor-based pagination everywhere"	When datasets are under 1000 items

---

Core Design Principles

REST Fundamentals

• Use nouns for resources, not verbs: GET /users not GET /getUsers
• Use plural for collections: /users, /posts
• Represent relationships hierarchically: /users/{id}/posts
• Use query parameters for filtering: GET /users?role=admin&status=active

HTTP Methods

Method	Purpose	Idempotent
GET	Retrieve	Yes
POST	Create	No
PUT	Replace	Yes
PATCH	Update	No
DELETE	Remove	Yes

GraphQL Principles

• Deprecation over versioning - Use @deprecated directive
• DataLoaders for N+1 prevention - Use batching when nested resolvers would create N+1 queries
• Query depth limiting - Set limits when the requested GraphQL contract needs them

---

Quick Reference Tables

HTTP Status Codes

Code	Meaning	When to Use
200	OK	General success
201	Created	Resource created (POST)
204	No Content	Success, no body (DELETE)
400	Bad Request	Invalid input
401	Unauthorized	Missing/invalid auth
403	Forbidden	Authenticated, no permission
404	Not Found	Resource doesn't exist
422	Unprocessable	Validation errors
429	Too Many Requests	Rate limit exceeded
500	Server Error	Unexpected error

Authentication Methods

Method	Best For	Complexity
API Key	Service-to-service, simple APIs	Low
JWT	Public APIs, stateless auth	Medium
OAuth2	Third-party integrations, "Login with X"	High

Pagination Strategies

Strategy	Best For
Offset/Limit	Small, static datasets
Cursor-based	Large, growing datasets (RECOMMENDED)
Keyset	Natural sort fields

---

Error Response Format

Standard structure:

{
  "code": "ERROR_CODE",
  "message": "Human-readable message",
  "details": {}
}

Error codes (use consistently):

• INVALID_REQUEST - Malformed request
• VALIDATION_ERROR - Field validation failed
• AUTHENTICATION_FAILED - Invalid credentials
• INSUFFICIENT_PERMISSIONS - Authorized but lacks permission
• RESOURCE_NOT_FOUND - 404
• RESOURCE_ALREADY_EXISTS - 409 on duplicate
• INTERNAL_SERVER_ERROR - 500

---

Common Pitfalls

Endpoint Explosion

Bad:  GET /users/admins, GET /users/active, GET /users/verified
Good: GET /users?role=admin&status=active&verified=true

God Endpoints

Bad:  GET /data?type=users&action=delete&id=123
Good: DELETE /users/123

Inconsistent Errors

Standardize error format across ALL endpoints.

Missing Pagination

Paginate collection endpoints when their requested contract needs bounded results; do not add pagination to small, intentionally complete collections.

Exposed Internals

Bad:  "ERROR: Unique constraint violation on users_email_idx"
Good: { "code": "VALIDATION_ERROR", "message": "Email already in use" }

Credentials in URLs

Bad:  GET /api/data?api_key=secret123
Good: GET /api/data (Authorization: Bearer <token>)

---

API Design Checklist

Apply the items required by the requested API contract:

• Consistent endpoint structure
• Clear error responses
• Proper status codes
• Pagination for bounded collection results
• Authentication/authorization for protected access
• Request validation
• Documentation (OpenAPI/GraphQL schema) when consumers need it
• Caching headers for REST responses whose caching behavior is part of the contract

---

Note: For project-specific API patterns, check the active repo/client instruction files, such as AGENTS.md, CLAUDE.md, README files, or local API docs.