Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Meza-dev/Ghostly/llms.txt

Use this file to discover all available pages before exploring further.

Ghostly is a local-first E2E testing engine that turns a plain-English goal into a fully executed Playwright browser flow — complete with self-healing selectors, live progress streaming, and artifacts like screenshots and video. Everything runs on your machine; no cloud required.

Quickstart

Install the CLI and run your first AI-assisted browser test in under five minutes.

Architecture

Understand how the CLI, API, runner, dashboard, and MCP server fit together.

CLI Reference

Full reference for every ghostly command: keygen, install, config, up, and update.

API Reference

REST endpoints for running flows, managing projects, and streaming live events.

How Ghostly works

1

Install the CLI

Install @ghostly-io/cli globally from npm and run ghostly keygen to create your local credentials.
npm install -g @ghostly-io/cli
ghostly keygen
2

Configure your AI provider

Connect an LLM — either an OpenAI-compatible HTTP endpoint or the local Cursor Agent CLI. Ghostly needs it to plan and heal test flows.
ghostly config
3

Start the engine

Run ghostly up to migrate the local SQLite database and start the API + dashboard on http://localhost:4000.
ghostly up
4

Run your first test

Open the dashboard, create a project, describe your goal in plain English, and watch the run execute live.

Key features

Natural Language Goals

Describe what the test should accomplish in plain English. The AI strategist generates, executes, and verifies the steps.

Self-Healing Selectors

When a selector breaks, the healer agent finds a working alternative automatically — no manual test maintenance.

Assist Memory

Successful flows are persisted and reused to make future runs faster and more reliable.

BYO LLM

Use any OpenAI-compatible endpoint (OpenAI, Ollama, OpenRouter, Anthropic proxy) or the local Cursor Agent CLI.

MCP Integration

Expose Ghostly tools directly inside Cursor and other MCP-compatible IDEs — run flows without leaving your editor.

TypeScript Client SDK

Integrate Ghostly programmatically with @ghostly-io/client — start runs, poll results, or stream events.
Ghostly is local-first by design. All cryptographic keys, run data, and LLM credentials stay on your machine and are never sent to external servers.

Build docs developers (and LLMs) love