I kept running into the same problem as a developer: API docs are often incomplete, outdated, or hard for both humans and AI to reason about. In practice, the real source of truth is usually the API surface itself, whether that's OpenAPI/Postman today or SDKs in the near future.
So I built Swytchcode, a tool that converts real API definitions into a structured spec that AI can use to generate integration code, workflows, and tests.
How it works today:
- Ingests OpenAPI & Postman specs (including large, real-world ones like Stripe)
- Normalizes endpoints, inputs, outputs, and error responses into a single spec (we call it a Wrekenfile)
- Uses that spec for AI-powered discovery, code generation, and workflow construction
What's coming next:
- Direct SDK parsing (Go, TypeScript, Python) to extract public methods and types
- Using SDKs as a source of truth when docs or specs are missing or outdated
- Deeper workflow and authentication discovery from real code paths
Why this approach:
- Docs drift over time
- Specs are better but often incomplete
- SDKs tend to reflect what actually works in production
- AI needs precise, machine-readable context to generate correct code
This is still early, but we're seeing steady developer usage, daily signups, and active conversations with API teams about integrating their APIs.
I'd love feedback from the HN community:
- Does this approach make sense?
- Where do you think SDK-based parsing could break down?
- What would you expect from a tool like this?
I kept running into the same problem as a developer: API docs are often incomplete, outdated, or hard for both humans and AI to reason about. In practice, the real source of truth is usually the API surface itself, whether that's OpenAPI/Postman today or SDKs in the near future.
So I built Swytchcode, a tool that converts real API definitions into a structured spec that AI can use to generate integration code, workflows, and tests.
How it works today: - Ingests OpenAPI & Postman specs (including large, real-world ones like Stripe) - Normalizes endpoints, inputs, outputs, and error responses into a single spec (we call it a Wrekenfile) - Uses that spec for AI-powered discovery, code generation, and workflow construction
What's coming next: - Direct SDK parsing (Go, TypeScript, Python) to extract public methods and types - Using SDKs as a source of truth when docs or specs are missing or outdated - Deeper workflow and authentication discovery from real code paths
Why this approach: - Docs drift over time - Specs are better but often incomplete - SDKs tend to reflect what actually works in production - AI needs precise, machine-readable context to generate correct code
This is still early, but we're seeing steady developer usage, daily signups, and active conversations with API teams about integrating their APIs.
I'd love feedback from the HN community: - Does this approach make sense? - Where do you think SDK-based parsing could break down? - What would you expect from a tool like this?
Happy to answer any technical questions.
Website: https://swytchcode.com Playground with Stripe (demo): https://playground.swytchcode.ai