Home/Docs/Getting Started

Getting Started with OutcomeDev

This guide will walk you through your first interaction with OutcomeDev.

1. Create a Task

The core unit of work in OutcomeDev is a Task. A task represents a single outcome you want to achieve.

  1. Navigate to the Home page.
  2. In the prompt box, describe what you want to build.
    • Example: "Create a landing page for a coffee shop with a contact form."
  3. (Optional) Click Options to configure settings like:
    • Install Dependencies: Checked by default. Uncheck if you want a faster start and don't need new packages.
    • Max Duration: How long the agent should work.

2. Choose an Agent

OutcomeDev supports multiple AI agents. Each has different strengths. Learn more in our Models & Agents Guide.

  • Claude (Anthropic): Best for complex coding tasks and large context.
  • Codex (OpenAI): Great for reasoning and following strict instructions.
  • Gemini (Google): Excellent for multimodal tasks (images, design).

Select the agent that best fits your task from the dropdown menu.

3. The Workspace

Once you create a task, you'll be taken to the Task Details page.

  • Chat: Communicate with the agent. You can give feedback or ask for changes.
  • Preview: See your web app running live in a secure sandbox.
  • Code: View the file changes and diffs generated by the agent.
  • Terminal: Monitor the logs and commands being executed.

4. Review & Iterate

The agent will work towards your outcome.

  • If the Preview looks good, you're done!
  • If something is missing, type your feedback in the chat. The agent will update the code and the preview will refresh.

5. Save Your Work

OutcomeDev automatically creates a Pull Request (PR) for your task. You can view this PR on GitHub to merge the changes into your main repository.