Quickstart

This guide walks you through creating your first agent in SuperAgent and having a conversation with it. By the end, you will have an agent running in a container, ready to take on tasks.

1. Open SuperAgent

Launch the desktop app or navigate to your SuperAgent server in a browser (default: http://localhost:47891).

On first launch, SuperAgent will check that a container runtime (Docker, OrbStack, or Podman) is available and pull the agent container image. This may take a minute or two the first time.

2. Configure your API key

Before creating an agent, make sure your Anthropic API key is configured. Open Settings (gear icon in the sidebar) and enter your key under the API Keys section. If the key was set via the ANTHROPIC_API_KEY environment variable, it will already be detected.

3. Create a new agent

From the main screen, you will see a text input where you can describe what you want your agent to do. This is the fastest path to creating an agent:

1. Type a prompt describing the agent's purpose -- for example, Summarize my unread emails every morning or Help me manage my GitHub issues.
2. Click Create Agent (or press Cmd+Enter / Ctrl+Enter).

SuperAgent automatically generates a name for the agent based on your prompt, creates the agent, and starts your first session.

Alternatively, you can create an agent from a template or import one from a skillset. These options appear below the main prompt input.

4. Your first conversation

Once the agent is created, you land in a chat session. Behind the scenes, SuperAgent spins up a container for the agent and sends your initial message.

You will see the agent begin working. It reads your message, thinks through the task, and uses its available tools -- running shell commands, reading and writing files, browsing the web, or calling connected services. Each tool invocation appears inline in the conversation so you can follow along.

Try sending a follow-up message to refine the task, ask a question, or give the agent additional context. The conversation works like a chat -- the agent maintains context across the entire session.

5. Choose a model

By default, new agents start their first session on Claude Opus for the strongest reasoning. You can switch models per-message using the model selector in the message composer. The available model families are:

• Opus -- Most capable; best for complex, multi-step tasks.
• Sonnet -- Balanced speed and capability; good default for everyday work.
• Haiku -- Fastest and most affordable; suitable for simple, high-volume tasks.

The model you select is remembered for the session, so subsequent messages continue using it unless you change it again.

6. Explore the agent home

When you navigate back to your agent (click its name in the sidebar), you land on the Agent Home page. From here you can:

• Start a new session -- Type a new prompt to begin a fresh conversation.
• Review past sessions -- All previous sessions are listed with timestamps and summaries.
• View dashboards -- If the agent has created any dashboards, they appear as cards.
• Check scheduled tasks -- Any recurring or future tasks the agent has set up.
• Manage connections -- See which connected accounts and MCP servers are available to the agent.

7. Configure agent settings

Click the Settings icon on the agent home page to access agent-level configuration:

• General -- Rename the agent, export it, or delete it.
• Secrets -- Add environment variables (API keys, tokens) that are injected into the agent's container.
• Agents -- Control which other agents in your workspace this agent is allowed to invoke (multi-agent policies).

What to try next

• Connect an account -- Go to the global Connected Accounts settings and link a service like Gmail, Slack, or GitHub. Then map it to your agent so it can act on your behalf.
• Schedule a task -- Ask the agent to perform a task on a recurring schedule. It will use the built-in scheduling tool to set up a cron job. See Scheduled Tasks.
• Build a dashboard -- Ask the agent to create an interactive dashboard for you. See Dashboards.
• Install a skillset -- Browse and install community skillsets to give your agent pre-built capabilities. See Skillsets.
• Read the core concepts -- Understand the key abstractions behind agents, sessions, containers, and tools.