Deterministic · No model calls · CLI as a skill
You test AI code with AI. This doesn't.
api-doctor scans AI generated code for bad api integrations. Deterministic AST rules. Not a prompt. Same input, same output, every time.
npx @api-doctor/cli .
Open SourceChecks integrations for
Run it. Get a report. Feed it back.
1
Scan
npx @api-doctor/cli .
Detects which SDKs you're using. Checks your code. Prints a report.
2
Install as an agent skill
npx @api-doctor/cli install
Your agent picks up the report and fixes findings on its own. Works with all coding agents.
3
Done
No tokens burned. No model calls. 100% deterministic.
Why determinism matters
AI Tests AI
- You write code in Claude Code
- You ask Claude to write tests
- If Claude hallucinated the code, the tests probably check for the wrong thing
- You have zero visibility into what's actually broken
api-doctor is different
- Hardcoded AST rules extracted from provider documentation
- Same rules, same output, every time
- Not a model. Not a prompt. Code that checks code.
- Rules are open source and verifiable on GitHub
The result
- Catch API integration bugs BEFORE your agent commits
- No surprise failures in production
- Deterministic = trustworthy
Auto fix every PR
api-doctor watches the moment the code gets written:
- implement integrations based on provider's best practices
- update with new SDK and API versions
Supported providers