> ## Documentation Index > Fetch the complete documentation index at: https://docs.stagewise.io/llms.txt > Use this file to discover all available pages before exploring further. # Quickstart — Your first prompt with stagewise > Connect a workspace to the stagewise agentic IDE and send your first AI coding agent prompt. Workspaces, @mentions, slash commands, and diff review in under 5 minutes. Connect a workspace and send your first prompt in under 5 minutes. ## Before you begin * [stagewise installed](/getting-started/installation) and launched * A project folder on your machine (any language, any framework) * Optional: signed in to your stagewise Account for instant model access. If you skipped authentication, make sure you've [imported your API keys](/reference/byok-setup) or [connected a local model](/reference/custom-providers) ## Step 1: Connect a workspace A workspace is any folder you want your AI coding agent to access. 1. Click the workspace panel in the sidebar 2. Click **Add workspace** and select a folder 3. The agent can now read, write, and modify files in that folder Connect multiple workspaces to let agents work across your full stack. [Learn more about workspaces →](/core-concepts/workspaces) Workspace selector showing worktree and branch options for connecting a project folder *Choose an existing workspace or create a worktree/branch-backed workspace before prompting the agent.* Workspaces are not sandboxed. The agent operates directly on your filesystem. Every change is tracked so you can review or undo anything. ## Step 2: Open the chat The chat panel is where you talk to your agent. You can: * **Type instructions** — Describe what you want in natural language * **@mention files** — Reference specific files for the agent to work on * **@mention tabs** — Attach browser tabs so the agent sees what you see * **Select DOM elements** — Click elements in the browser to attach them as visual context * **Use slash commands** — Type `/` to access skills, plugins, and agent features Chat input with image, browser tab, and DOM element context attached to a prompt *Attach context directly in the chat input so the agent knows which files, tabs, screenshots, or elements matter.* ## Step 3: Send your first prompt Type a prompt and press Enter. Start with something exploratory: > *"Summarize the structure of this project and suggest three improvements."* The agent reads your files, analyzes the project, and responds with a plan. You can then ask it to implement the suggestions. ## Step 4: Review changes Every file edit the agent makes is tracked. After the agent modifies code, you'll see an inline diff summary in the chat: * **Accept** changes you want to keep * **Reject** changes to revert them * **Undo** or **redo** entire tool calls Open the **Diff Review** page to see all pending changes across files. [Learn more about diff review →](/core-concepts/diff-review) ## Add context and commands Use the chat input for mentions and slash commands. Use the browser toolbar or hotkey for DOM selection. | Action | How to do it | What happens | | --------------------- | ------------------------------------------------------- | ---------------------------------------------- | | Mention a file | Type `@` in the chat input, then choose a file | Attaches the file as context | | Mention a browser tab | Type `@` in the chat input, then choose a tab | Attaches the page as context | | Run a slash command | Type `/` in the chat input, then choose a command | Opens skills, plugins, and agent actions | | Select a DOM element | Click the DOM selector button or press `⌘ I` / `Ctrl+I` | Attaches a debug file for the selected element | ## Write useful prompts Give the agent a specific target, expected behavior, and any constraints. Good prompts include: * The file, component, or page you want changed * The exact behavior you expect * Screenshots, browser tabs, files, or DOM selections when visual context matters Example: > *"Add input validation to the login form and show error messages below each field."* Start a new chat for each distinct task. Use the DOM context selector when a visual change is easier to point at than describe. ## What's next Understand the agent loop and how AI coding agents process your requests. Connect multiple workspaces for full-stack agent workflows. Teach agents project-specific patterns and integrate your stack. Use existing provider API keys.