API Design Principles Check your API →
Runs in your browser · your spec is never uploaded

Score your API design in
seconds.

Paste an OpenAPI spec — get a graded report against the 10 principles, a prioritized list of fixes, and the reasoning behind every rule. No login · no config · no CI

Score my spec → Read the 10 principles
openapi.yaml Upload 
# openapi.yaml — paste yours, or load a sample
openapi: 3.0.3
info:
  title: Petstore
  version: 1.0.0
paths:
  /createUser:      # ← verb in path
    post:
      summary: Create user
      responses:
        '200':
          description: ok
  /users/{id}:
    get:
      parameters:
        - in: query
          name: api_key   # ← secret in query
OpenAPI 3.x · YAML / JSON
API design score
B+84 / 100
2 errors 3 warns 1 info
Naming0
HTTP semantics0
Errors0
Versioning0
Security0
100 public APIs scored Open-source scoring core No display ads, ever
The framework

Ten principles, three things they add up to

Human developers read docs and interpret edge cases. AI agents can't — so the principles cluster around three outcomes: an API an agent can find, call safely, and rely on.

Discoverable

Runtime discovery via OpenAPI, llms.txt, or MCP — agents find what they can call.

Safe to call

Scopes, least privilege, and permission boundaries that hold up to autonomous use.

Consistent

Predictable schemas, structured retryable errors, and consistent pagination.

Read all 10 principles →
Why generic OpenAPI isn't enough

A valid spec is not an agent-ready spec.

An OpenAPI document can pass every linter and still overwhelm an agent: 300 near-identical CRUD endpoints, vague descriptions, and no high-level capabilities to select from. Validity measures syntax. Agent-readiness measures whether a model can choose and call the right operation safely.

Read: OpenAPI → MCP guide
Trust layer

Benchmarks & buyer's guides

Independent, versioned, and separated from sponsorship — the neutral referee for agent-ready API tooling.

Index

Public API Agent-Readiness Index

100 public APIs scored on the same transparent rubric.

 Benchmark

Annual MCP Generation Benchmark

How spec-to-MCP tools compare on real specs.

 Guide

Best MCP Gateways (2026)

Buyer's guide with disclosed affiliate relationships.
Owned audience

Get the agent-ready API briefing

Monthly: new index scores, MCP standard changes, and the fixes that move the needle. No spam.

Double opt-in · unsubscribe anytime.

Partner & referral disclosure. Some buyer's guides contain affiliate or referral links, always labelled. Sponsorship never influences benchmark scores or index rankings. Our scoring methodology is public and versioned. — Editorial policy