← Back to leaderboard
Plaid API
63
C
Plaid scores 63, a C. The top-ranked spec in this analysis. Semantics, intent, and pagination all score 100. The spec is well-organized and operations are clearly named. But two gaps are sharp. Error handling scores 0: 329 of 330 operations do not document error responses. For a financial API where the difference between "ACCOUNT_NOT_AUTHORIZED" and "ITEM_LOGIN_REQUIRED" determines whether an agent retries or asks the user to re-authenticate, this is a meaningful blind spot.
Category breakdown
Examples70
Semantics100
Intent100
Error Handling0
Parameters29
Pagination100
Key findings
- 2,107 of 2,254 schemas lack examples
- 329 of 330 operations do not document error responses
- No RFC 9457 Problem Detail format detected
- Semantics, intent, and pagination all score 100, the highest of any API in this study
- 13 of 330 operations lack request and response examples
What Plaid should fix
- Document error responses across all operations. For financial APIs, error specificity is the difference between a recoverable retry and a stalled user flow.
- Add examples to the 2,107 schemas missing them. Most are nested response objects where examples teach structure faster than reading schemas.
- Add descriptions to parameters. Plaid's domain terms (`access_token`, `cursor`, `count`) all have specific meanings that are not obvious.
How does your API score?
Run AgenticScore on your own OpenAPI spec — find issues before LLMs do.
npx agenticscore score ./openapi.yaml
Install the CLI →