moonpeak

Testing Framework

E2E testing

Playwright tests against the inspector, in ChatGPT and Claude style hosts.

View .md Edit on GitHub

npx playwright install chromium   # once
moonpeak test --e2e

moonpeak starts the inspector for your project and runs your Playwright config. Tests import from moonpeak/test:

import { test, expect } from "moonpeak/test";

test("renders in Claude dark mode", async ({ inspector }) => {
  const view = await inspector.open({ tool: "show-moon", sim: "show-moon#full-moon", host: "claude", theme: "dark" });
  await expect(view.getByRole("heading", { name: "Full Moon" })).toBeVisible();
});

test("asking posts a message", async ({ inspector }) => {
  const view = await inspector.open({ tool: "show-moon", sim: "show-moon#full-moon" });
  await view.getByRole("button", { name: "Ask about this night" }).click();
  await expect(inspector.sentMessages()).toContainText("full moon");
});

test("tool output", async ({ mcp }) => {
  expect(await mcp.call("show-moon", { date: "2026-11-24" })).toHaveStructuredContent({ phase: "Full Moon" });
});

`inspector` fixture

Member
`open({ tool, sim, host, theme, device, mode })`	opens the inspector with those settings and returns a frame locator for the View once it has rendered
`view()`	the current View's frame locator
`sentMessages()`	locator for messages the View posted
`modelContext()`	locator for what the View told the model
`setTheme(t)`, `setDisplayMode(m)`	change host settings without reloading the View
`page`	the Playwright page

`mcp` fixture

Member
`call(name, args)`	call a tool, returns the `CallToolResult`
`listTools()`	the server's tools, with `_meta`
`client`	the raw MCP client

Matchers

toHaveStructuredContent(partial), toBeError() and toHaveTextContent(text), using the same partial matching as simulations.