---
title: "Quickstart"
description: "Sign up, create a project, run your first Generate session and hand the specs to your coding agent."
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.myspec.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Quickstart

This guide takes you from a new account to a finished set of spec files.
It takes about ten minutes, most of it spent answering the interview.

## 1. Create your account

1. **Sign up**

   Go to [app.myspec.dev](https://app.myspec.dev) and register with your email
   address, or sign in with Google, GitHub, GitLab, Linear or Atlassian.
   If you register with email, confirm your address from the email we send you.
2. **Set up your workspace**

   Create a workspace (an organization) for your team, or choose
   **Skip and set one up for me**. If you were invited to an existing workspace,
   pick it instead.
3. **Tell us your role**

   Pick the role closest to how you work so MySpec can tailor its questions.
   You can skip this and change it later.

## 2. Create a project

1. On the **Projects** page, select **New Project**.
2. Give the project a name and a description.
3. Choose a project type:

| Type | Use it when |
|---|---|
| **Greenfield** | You are starting from scratch. Describe what you want to build. |
| **Brownfield** | You want to change an existing codebase. MySpec explores the code first, so the description is optional. |
| **Auto** | You are not sure yet. MySpec decides based on the workflow you pick. |

4. Select **Create Project**.

> **Brownfield projects read your code**
>
> For a Brownfield project, MySpec needs read access to a folder on your machine
> through the local MCP server's `reverse` command. See
> [Local MCP server](/mcp/local-server/#share-a-local-folder-for-brownfield-projects).

## 3. Run a Generate session

A session is a conversation with the AI that produces or edits spec files.

1. In your project, start a new chat and choose **Generate**.
2. Answer the interview. Questions are often multiple choice; pick **Other**
   to write your own answer.
3. MySpec writes each spec file in turn. For each one you can accept it, ask
   for changes, or review the AI's own review notes first.
4. Repeat until the last stage is done, then complete the session.

The default workflow runs these stages:

| Stage | What you get |
|---|---|
| Interview | Questions about your project's goals, users and constraints. No file yet. |
| Constitution | `constitution.md`: guiding principles and non-negotiable rules. |
| Requirements | `requirements.md`: user stories plus functional and non-functional requirements. |
| Solution | `solution.md`: architecture, data model and key design decisions. |
| Tasks | `tasks.md`: a concrete, ordered work plan ready for implementation. |

> **Note**
>
> Type `/` in the message box to see commands. For example, `/btw` asks a side
> question without changing the session, and `/complete` finishes the session.

## 4. Build from your specs

When the session is complete you can:

- **Download** the spec bundle as a ZIP file.
- **Open it in a vibe-coding tool** such as Lovable, Bolt or Replit to scaffold
  a prototype.
- **Generate an AI prompt** for Cursor, GitHub Copilot or Claude.
- **Connect your coding agent** to MySpec over MCP, so it can read the latest
  specs directly and keep them up to date.

- [Connect your coding agent](/mcp/) — Set up the MySpec MCP server in Claude, ChatGPT, Cursor or your editor.

Source: https://docs.myspec.dev/get-started/quickstart/index.mdx