Quick Start

Quick Start Guide

Get up and running with Kollab in under 5 minutes. This guide will walk you through the basics.

1 Install Kollab

Install Kollab with uv. It fetches Python 3.12 for you if you don't have it:

# Install with uv
uv tool install kollab

Other ways: Homebrew on Apple Silicon Macs, an install script that needs no Python, pipx or pip. See Installation.

2 Launch Kollab

Start the application by running:

# Start Kollab
kollab

You'll see the Kollab welcome screen with the input prompt at the bottom. With no provider set up yet, Kollab opens the setup wizard by itself. Prefer a browser? kollab --web-ui opens the same agents there.

3 Set Up a Model With /setup

Kollab needs a model to talk to. /setup (also /onboard and /wizard) is a guided wizard, so you need no environment variables or hand-edited JSON:

  1. Pick a provider: Anthropic, OpenAI with an API key, OpenAI with your ChatGPT sign-in, Google Gemini, OpenRouter, or Custom / Local for any OpenAI-compatible endpoint or a local model.
  2. Add a key and confirm the endpoint. The key is optional for a local server.
  3. Choose a model from the list, or type a model id.
  4. Review. Press t to test the connection, then Enter to save and switch to it.

Prefer environment variables? Kollab also picks up standard provider keys such as ANTHROPIC_API_KEY. Add the export to your ~/.zshrc or ~/.bashrc to persist it across sessions.

4 Start Chatting

Type your message and press Enter to send. Kollab supports:

  • Multi-line input (Shift+Enter for new line)
  • Conversation history (scroll to review)
  • Real-time streaming responses
  • Syntax highlighting for code

5 Use Slash Commands

Type / to see available commands:

/save

Save conversation to file

/branch

Fork conversation from a specific message

/matrix

Enter the Matrix effect

/help

Show all commands

6 Try Pipe Mode

Use Kollab in scripts with pipe mode for quick queries:

# Single query mode
kollab your-query-here

Or pipe from stdin:

echo "What is a REST API?" | kollab -p

7 Use Agents

Agents are custom configurations with specialized system prompts. Create agents in your project's .kollab/agents/ directory, or globally in ~/.kollab/agents/:

# Create an agent directory
mkdir -p .kollab/agents/code-reviewer

# Add a system prompt
cat > .kollab/agents/code-reviewer/system_prompt.md << 'EOF'
You are a code reviewer. Focus on:
- Code quality and best practices
- Potential bugs and edge cases
- Performance considerations
EOF

Switch to your agent:

/agent code-reviewer

8 Load Skills

Skills are markdown files that add specific instructions to your agent. Create a skill:

# Add a skill to your agent
cat > .kollab/agents/code-reviewer/security-check.md << 'EOF'
## Security Check Skill

When reviewing code, also check for:
- SQL injection vulnerabilities
- XSS attack vectors
- Hardcoded secrets or credentials
- Input validation issues
EOF

Load and unload skills dynamically:

/skill security-check# Load the skill
/skills# List available skills
/skills unload security-check# Unload the skill

9 Explore Plugins

Extend Kollab with plugins. Kollab loads plugins from the project's plugins/ directory and from the installed package; base classes live in the kollabor_plugins package:

# my_plugin.py
from kollabor_plugins import BasePlugin

class MyPlugin(BasePlugin):
    def register_hooks(self):
        self.register_hook('pre_user_input', self.my_handler)

    async def my_handler(self, context):
        # Your custom logic here
        return context

You're All Set!

You've completed the quick start guide. Here are some resources to go deeper: