APIForge

API quality score for your OpenAPI or Swagger spec

Paste a spec URL and get a REST API design score from 0 to 100. Seven fixed rules cover path naming, HTTP verbs, structure, consistency, versioning, error responses and documentation. Each lost point traces back to a route.

No spec handy?
  • Rule-based, not AI
  • 7 pillars
  • OpenAPI 3.x
  • Swagger 2.0

How the score is calculated

Every operation is scored on its own, the results are averaged per rule, and the seven rule scores are added up. The weights below sum to 100.

CategoryMax pointsWhat lowers it
Path naming20camelCase in a path, underscores in a segment, and verb segments such as get, create, update, delete, list or save.
HTTP verb usage20A GET with a request body, and action words (create, update, delete, remove, get, fetch, add, save) in the path.
Path structure and depth15Paths deeper than 3 segments. The issue list only flags paths deeper than 4.
Consistency15Duplicate method and path pairs, and singular and plural forms of the same word across paths (/user and /users).
Versioning10No path with a segment such as /v1 or /v2. One path is enough for full points.
Error handling coverage10The share of operations with no 4xx, 5xx or default response. If no operation declares any responses, this rule gets 2 of 10.
Operation documentation10The share of operations with neither a summary nor a description.

The score reads the design of the extracted operations only. For the reasoning behind each rule, see the API quality checklist.

Related tools and guides

Frequently asked questions

Seven rules, each worth a fixed number of points: path naming 20, HTTP verb usage 20, path structure and depth 15, consistency 15, versioning 10, error handling coverage 10, operation documentation 10. APIForge scores every operation, averages the result per rule and adds the seven numbers, rounded to one decimal. There is no network call and no model in the calculation.

No. The same spec always gets the same score because the rules are plain code. AI features in APIForge, such as payload and test generation or the workspace assistant, run only when you ask for them and do not change the score.

100 means none of the seven rules found a problem. There is no industry benchmark behind the number, so treat it as a way to compare a spec with its own earlier version. The API quality checklist guide scores a sample spec at 58.7 before fixes and 100 after.

Start with the pillar that lost the most points. Use kebab-case plural nouns in paths, move verbs out of URLs, flatten paths deeper than three segments, add a /v1 segment, declare 4xx and 5xx responses on every operation, and give each operation a summary. The workspace shows a fix preview for renames such as POST /getUserOrders to GET /users/{id}/orders.

No. It scores the design of the operations it can extract, not the correctness of the file. A spec can score 100 and still contain schema mistakes. Use the OpenAPI validator to see whether the file parses and its references resolve.

Authentication and security schemes, request and response schema quality, pagination, rate limiting and examples are outside the seven rules. A high score says the URLs, methods, errors and documentation are tidy. It does not say the API is secure or correct.

The versioning rule gives the full 10 points if at least one path has a segment like /v1 or /v2, and 1 point if none does. Header or media-type versioning is not detected, so a spec that versions that way still loses the points. The API versioning guide compares the strategies.

The parse request fetches the spec on the server, returns the operations to your browser and does not save the spec. APIForge is free, needs no account and is open source under the MIT license.