AI Integration

Glubean’s SDK is designed from the ground up to be understood and written by AI assistants like Cursor, GitHub Copilot, and ChatGPT.

Because Glubean uses standard TypeScript, explicit types, and rich JSDoc with @example tags, AIs rarely hallucinate APIs or make syntax errors.

The AI Workflow

Instead of manually constructing HTTP requests in a GUI or writing boilerplate fetch code, use your AI to do the heavy lifting.

1. Provide the spec Paste your OpenAPI / Swagger spec (or even just a curl command or endpoint description) into your AI chat window.

2. Ask for a test Prompt the AI using natural language:

“Generate Glubean tests for the user CRUD endpoints. Create a user, fetch it to verify, update the name, and then delete it.”

3. Run it Save the generated code into a file in your explore/ folder (e.g., explore/user-crud.test.ts). Click the ▶ play button in the gutter.

You’ve gone from spec to running tests in seconds.

Better context: glubean context

To get the most accurate tests from your AI, it needs to understand the patterns already established in your project, and what the SDK is capable of.

The CLI provides a command to bundle all this context together:

glubean context --openapi openapi.json

This command generates a .glubean/ai-context.md file containing:

• The Glubean SDK quick reference
• Existing test patterns found in your current project
• A list of uncovered endpoints from the provided OpenAPI spec

How to use it: At the start of a new chat session (or as a rule in Cursor), tell your AI to read .glubean/ai-context.md before generating new tests. This ensures the output matches your specific assertions, ctx.vars usage, and project structure.