Skip to main content
This guide walks you through creating your first project, running a task, and reviewing the result. You’ll have an AI agent writing code in under 5 minutes.

​
1. Create a project

1

Open Claude Board

Launch the desktop app. If this is your first launch, the setup wizard will guide you through initial configuration.
2

Click 'New Project'

Enter a project name, slug, and the working directory — the local path where Claude will read and write code.
Point the working directory at an existing repository for best results. Claude works best when it can read your existing code structure.
3

Configure permissions

Choose a permission mode:
Claude runs with --no-input, so it cannot ask for permission interactively. Use “Auto Accept” for best results.

​
2. Create and run a task

Click 'New Task'

Write a clear, specific description. The more detail you provide, the better Claude performs.
Good:
Too vague:

Configure the task

Start the task

Drag the task to In Progress or right-click and select “In Progress”. Claude spawns immediately, creates a branch, and starts coding.

Watch the live terminal

Click the terminal icon on the task card to watch Claude work in real-time. You’ll see:
  • Read — files Claude is examining
  • Write/Edit — code changes being made
  • Bash — commands being executed (tests, git, etc.)

Review the result

When Claude finishes, the task moves to Testing. Click the review button to:
  • Approve — moves to Done, optionally creates a PR
  • Request Changes — provide feedback, Claude iterates automatically

​
3. Level up

Now that you’ve run your first task, explore these features to get more out of Claude Board:

Orchestration

Set up dependency graphs and run tasks in parallel waves.

Auto-Queue

Configure hands-free execution with concurrency limits.

Context Snippets

Inject project rules and coding standards into every prompt.

Git Automation

Auto-create branches and PRs per task.