Appear

Appear has wound down. The founders are now building ctxpipe.ai — the context layer for AI engineering agents.

OpenAPI validator

Spectral, Swagger Parser, and OAS Zod, plus a deterministic documentation checklist. Static analysis of the spec — not a test of agents. No account, no model.

Upload your OpenAPI 3.x spec

Drag and drop your JSON or YAML file here, or click to browse

Refreshing your page will discard the data. We don't store your spec.

Maximum file size: 4.5 MB

No validation results yet

Upload an API spec to see how different validators interpret and validate it.

What the Agent-Ready Score measures

This is static analysis of the OpenAPI document. Invalid specs fail anything that consumes the file. Everything else is a checklist of documentation that generated tools typically need — not a prediction that agents will succeed. Spectral, Swagger Parser, and OAS Zod run first. Same file, same score.

Six scoring dimensions

Foundational Compliance25%
The spec should parse. Broken $refs and invalid schemas stop anything that consumes the document, including generated tools.
Semantic Richness20%
Descriptions are what a generated tool has to go on. Missing prose is a documentation gap, not proof of a wrong call.
Agent Usability20%
operationIds are often used as tool names. This looks for them, plus pagination and idempotency hints in the document — not at runtime.
AI Discoverability15%
Examples and tags are useful hints in the spec. They are not a substitute for a working integration.
Security10%
Documented auth schemes tell a consumer what to send. This does not test whether authentication works.
Error Recoverability10%
Documented error schemas help a caller interpret failures. This checks the spec, not live Retry-After headers.

Frequently asked questions

What is an Agent-Ready Score?

A rule-based 0–100 checklist of an OpenAPI document: parseability, descriptions, operationIds, examples, documented auth, and documented errors. It is static analysis of the file, not a test of agents in production. Grade A/B (80+) means strong coverage of this checklist — not that the API is “agent ready”.

Does this tool use AI to calculate the score?

No. Same spec, same score, every time. The number is coverage of documented checks. It does not predict how an agent will behave.

Why is my score low when the API works for human developers?

Humans infer intent. Generated tools mostly cannot. Missing operationIds, examples, error schemas, or descriptions are easy for people and leave tool-calling setups with less to work with. That is a documentation gap, not proof the API is broken.

What is the difference between Technical Validation and the Agent-Ready Score?

Technical Validation (Spectral, Swagger Parser, OAS Zod Validator) catches structural errors. The checklist also flags documentation that is optional for humans but commonly used when a spec is turned into tools. A valid spec can still score poorly. A high score is not a certificate.

Which OpenAPI versions are supported?

OpenAPI 3.0.x, 3.1, and 3.2 in YAML or JSON. OpenAPI 3.2 is checked by OAS Zod Validator; Spectral and Swagger Parser are skipped until those engines support 3.2.

Who built this, and what happened to Appear?

This validator was built at Appear and remains free and open source. Appear has wound down. The founders now build ctxpipe.ai, a self-learning context layer for AI engineering agents.