Glubean for VS Code

Developer-owned API verification — code-first in TypeScript, accelerated by AI.
Verify REST, GraphQL, gRPC, browser, and anything Node.js supports.

verification · code-first · data-driven · result viewer · AI-native · CI-ready

💬 Ask Glubean AI anything — setup help, feature questions, comparisons

Demo

Why Glubean?

• Not a REST client — Write verification code that runs in CI, not click-through workflows that break.
• Zero to running in 10 seconds — Create a .test.js, click play. No setup, no config, no npm install.
• See every detail — Result Viewer shows traces, assertions, response schema, headers — all in one place.
• AI writes your tests — MCP server + skill + schema inference = AI understands your API and writes verification code.
• Data-driven at scale — test.each + test.pick with JSON/YAML/CSV data files, CodeLens shows every case.

Quick Start

Prerequisites: Node.js 20+ (download)

1. Install — from the VS Code Marketplace, Open VSX, or download a VSIX for Cursor / VSCodium.

2. Try it — create a .test.js file, type gb-scratch to insert a starter test, and click Play:

import { test } from "@glubean/sdk";

export const getProducts = test("get-products", async (ctx) => {
  const res = await ctx.http.get("https://dummyjson.com/products");
  ctx.expect(res).toHaveStatus(200);

  const data = await res.json();
  ctx.expect(data).toHaveProperty("products");
  ctx.log("total", data.total);
});

No npm install, no package.json, no setup. The response opens in the Result Viewer right beside your code.

3. Create a project — when you're ready for the full experience, run npx glubean init in your terminal to scaffold a project with environments, secrets, CI config, and more.

4. Set up AI — if you use Claude Code, Cursor, or Codex, connect glubean's MCP server and skill:

npx glubean config mcp       # AI can discover, run, and diagnose tests
npx skills add glubean/skill  # AI learns glubean patterns to write tests

Note: SDK reference docs are bundled with the skill — no separate download needed.

Now your AI agent can write verification tests, run them via MCP, read structured failures, and fix until green — without leaving the chat. Learn more →

Scratch mode vs Project mode — The single-file experience above is scratch mode: great for trying things out and quick API checks. For .env files, secrets, test.each / test.pick, CI upload, and project-level configuration, create a project with npx glubean init. See limitations →

Features at a Glance

Each feature is documented in detail at docs.glubean.com.

Documentation

• Quick Start — install and run your first test
• Writing Tests — CodeLens, snippets, data-driven patterns
• Running Tests — play buttons, Test Explorer, Glubean Panel, rerun failed
• Result Viewer — traces, assertions, history, jump to source
• Environments & Secrets — .env files, secrets, status bar switcher
• AI Integration — MCP server, skill, AI authoring loop
• Debugging — breakpoints, step-through, Debug Console
• Diagnostics — troubleshooting, common problems
• Commands & Settings — full reference
• Limitations — trade-offs and known limitations

How It Works

1. Discovery — static regex scans *.test.{ts,js,mjs} for test(), test.each(), test.pick() patterns.
2. Display — each test becomes a TestItem with play buttons in the gutter and Test Explorer.
3. Execution — clicking play runs the test via the bundled @glubean/runner. No external process needed.
4. Results — the runner streams events; the extension writes .result.json to .glubean/results/ and opens in the Result Viewer.
5. History — prev/next navigation across runs, rerun failed tests with one click.

License

MIT