specscore scores your OpenAPI spec across 6 dimensions and gives it a letter grade — so you know exactly what to fix before connecting it to an LLM.
Most OpenAPI specs are written for humans. AI agents have different needs — and most specs fail silently.
Run against any OpenAPI 3.x spec and get a scored report in milliseconds. No account, no upload, no waiting.
Dimensions are weighted for what LLMs and agents actually need: semantic operation IDs, error schemas, intent descriptions.
Every failing check includes the exact JSON path and a plain-English explanation so you know what to fix and where.
Structured exit codes (0 / 1 / 2) and --json output make it trivial to gate deploys on API quality.
Pure Python CLI. Runs locally, in Docker, or in any CI runner. Your spec never leaves your machine.
A single A–F grade tells the whole story. Drill down per-dimension to understand exactly where quality is lost.
clitic — CLI Intelligence & Compliance Tester — probes any command-line tool and scores it across 5 dimensions. Know exactly how well your CLI will work inside an AI agent loop before you ship.
Each dimension probes a different aspect of agent-friendliness by actually running the tool.
Each dimension targets a distinct quality axis that determines whether an AI agent can discover, call, and reason about your API.
No account. No config. Just point it at an OpenAPI spec.
Official release — same tool as on this site. One command and you are ready to score specs.
0
Score ≥ 60 — passing
1
Error reading or parsing the spec
2
Score < 60 — failing grade
Gate your CI pipeline on API quality:
specscore score spec.yaml || exit 1
Questions, feedback, or want to contribute? Reach out on any of these channels.
No clicks recorded yet.
localStorage · GA4 events also fired · Shift+S to close