Divio type: Tutorial
In this tutorial, you’ll install Spec Kitty and create your first feature specification.
Time: ~30 minutes Prerequisites: Python 3.11+, Git, an AI coding agent (Claude Code, Cursor, Gemini CLI, etc.)
Recommended: Install jujutsu (jj) for automatic rebasing and non-blocking conflicts. Spec Kitty will use jj when available. See the Jujutsu Workflow Tutorial for details.
Choose one install method:
pip install spec-kitty-cli
uv tool install spec-kitty-cli
Verify the CLI is available:
spec-kitty --version
Expected output (abridged):
Spec Kitty ... v0.11.0
Create a new project directory with the agent you plan to use:
spec-kitty init my-spec-project --ai claude
cd my-spec-project
Expected output (abridged):
OK Initialized Spec Kitty project
OK Generated agent commands in .claude/
Tip: Use spec-kitty init --here --ai claude to initialize the current folder.
Open your AI agent in this repository and run the specify command.
In your agent:
/spec-kitty.specify Build a tiny command-line task list app.
You’ll be asked a discovery interview. Answer each question until the command completes.
Expected results:
kitty-specs/###-task-list/spec.md (feature spec)main with the spec changesConfirm the feature directory exists:
ls kitty-specs
Example output:
###-task-list
If the command created a new worktree later in the workflow, it will appear here:
ls .worktrees
spec-kitty: command not found: Reopen your shell or reinstall via pipx or uv. Then rerun spec-kitty --version./spec-kitty.specify command available: Re-run spec-kitty init --ai <your-agent> or use spec-kitty agent context update-context to refresh agent context.WAITING_FOR_DISCOVERY_INPUT: The command is paused for your answers; provide the requested details and continue.Continue with Your First Feature for the complete workflow from specification to merge.
/spec-kitty.specify