Free tool
API Agent-Readiness Checker
Upload a spec, paste a URL, or try a demo. Public scans are free — results in seconds.
Drop your openapi.yaml / .json
OpenAPI 3.0 / 3.1 · max 5 MB
The URL must allow cross-origin requests (most public spec hosts do).
Score a sample spec
A small, well-formed Bookstore API (OpenAPI 3.1).
What you get
- Overall score & letter grade
- Top 5 issues + priority fixes
- Endpoint-surface & description review
- Auth, pagination & error analysis
- Embeddable readiness badge
Scoring runs locally in your browser — your spec is never uploaded to a server.
Sample result — api.acme.com
72
B− AGENT-READY
SCORE /100
SCORE /100
Semantic clarity 88
Naming consistency 82
Auth model 79
Pagination 76
Rate limits 64
Error handling 61
Least privilege 58
API surface size 54
Tool curation 38
Discovery 30
Top 5 issues & priority fixes
High
No agent-discoverable surface. No llms.txt, MCP manifest, or machine-readable capability map. Fix: Publish an llms.txt and group operations into capabilities.
High
312 raw CRUD endpoints, no curation. Surface is too large for reliable tool selection. Fix: Expose 12–20 high-level capabilities over the raw endpoints.
Med
41% of operations have empty descriptions. Agents can't disambiguate. Fix: Add one-line intent descriptions to every operation.
Med
Errors are unstructured strings. No machine-readable codes for recovery. Fix: Adopt a typed error envelope with retryable flags.
Low
Inconsistent pagination. Cursor and offset styles mixed across collections. Fix: Standardise on cursor pagination.
Sample responses — what an agent would receive
Upload or load a spec to preview the example response for each operation.
Want the full report? Get the complete 10-dimension breakdown + fix playbook as a PDF.
Scored with methodology v1.3 · scores are independent of any vendor.