Quickstart

This guide walks you through your first few minutes in Chatty the Lab: signing in, picking a workspace, running a chat against a model, and finding your way to the Studio to build a graph.

Run your first chat

1. Sign in. Open the app and authenticate with Google, email, or a magic link. The first time you sign in, a personal workspace is created for you automatically.

2. Pick a workspace. Use the WorkspaceSwitcher in the navbar to confirm you are in your personal workspace (or a team you belong to). Everything you create lands in the active workspace.

3. Add a provider key. A chat needs a model, and most models need a provider key. Go to Settings → Providers and add a key for at least one provider (OpenAI, Mistral, Gemini, HuggingFace, or NVIDIA Build). To use local models instead, install Ollama — no key required.

4. Start a chat. Open Chats and create a new session. You will see a sidebar of chat sessions and a chat area with a message list and an input box.

5. Pick a model. Use the ModelPicker to choose the model your chat should use. You can add more chats to the same session, each on a different model, to compare answers side by side.

6. Send a prompt. Type a message (or pick one from the PromptPicker) and send it. The response streams back into the chat area.

Tip

Want to compare models? Add several chats to one session, each bound to a different model, and run the same prompt against all of them. See Chats overview.

Build your first agent graph

Once you are comfortable chatting, head to the Studio to build a reusable agent workflow on a visual canvas. Drag nodes from the catalog, connect them, configure each one, and run the graph to watch live execution.

Build a graph Add an Input node, wire up the catalog, set exposed params, and run it.

Note

A graph you build in the Studio can later be bound to a chat, so it runs on every turn. See Binding a graph.