Introducing Kollabor - AI Collaboration for Everyone

Today we're launching Kollabor, an AI collaboration platform built on two simple principles: meet people where they work, and give them full control. We're not trying to replace your workflow—we're building AI that actually works with it.

Two Products, One Philosophy

Kollabor comes in two forms. Kowork is a desktop workspace built for voice-first interaction. Kollab CLI is a terminal-based AI assistant where everything has hooks. Both share the same core philosophy: local-first, extensible, and built for developers who care about their tools.

Kowork: Voice, Reimagined

The centerpiece of Kowork is The Orb—a global voice capture system that lives on your desktop. Hold Fn and speak from any application, and your words appear exactly where you need them. No switching windows, no context switching, no friction.

Built with Tauri and Vue 3, Kowork runs natively on macOS, Windows, and Linux. We use Rust for the backend to handle audio processing efficiently, and whisper.cpp for local transcription. Your recordings never leave your machine unless you want them to.

The app includes a full chat interface with support for Anthropic's Claude and OpenAI-compatible APIs. Tool calling lets the AI execute commands, read files, and interact with your development environment. Language Server Protocol integration brings IDE-quality code understanding right into your conversations.

Kollab CLI: Everything Has Hooks

For those who live in the terminal, Kollab CLI offers something different: an event-driven architecture where every interaction is hookable. We've built a plugin system with 30+ event types covering the entire application lifecycle.

Plugins can intercept user input, transform LLM requests, process responses, and even create custom UI elements. The hook system includes priority levels (SYSTEM, SECURITY, LLM, POSTPROCESSING, DISPLAY) so plugins can coordinate cleanly without conflicts.

Want to add custom logging? Hook into LLM_RESPONSE_POST. Need to approve certain tools? Hook into TOOL_CALL_PRE at SECURITY priority. Building a custom status widget? Hook into RENDER_FRAME. The architecture stays out of your way while giving you complete control.

MCP Integration

Both products fully support Model Context Protocol (MCP), the open standard for connecting AI assistants to external tools and data sources. MCP servers can provide tools, resources, and prompts to your conversations, and Kollabor exposes the full MCP event lifecycle for plugins to hook into.

Discover MCP servers, register their tools, and call them seamlessly—all with proper permission checks and error handling. The MCP_STATUS_PLUGIN shows connected servers and available tools right in your status bar.

Permission System That Doesn't Suck

Tool calling is powerful, but you need control. Kollab CLI includes a sophisticated permission system with four approval modes: CONFIRM_ALL (default), DEFAULT, AUTO_APPROVE_EDITS, and TRUST_ALL. Risk-based assessment categorizes tools by danger level, and inline permission prompts let you approve once, for the session, for the project, or deny entirely.

The permission system integrates at SECURITY priority in the event bus, so plugins can participate in approval decisions. Session-scoped approvals mean you don't get nagged repeatedly for the same tool.

Tmux Integration

For terminal power users, Kollab CLI includes deep tmux integration. Create new sessions, view existing ones, attach, kill—all from slash commands. The tmux plugin even offers a live modal that lets you watch your tmux sessions in real-time without leaving the chat.

Slash Commands with Subcommands

The command system is fully extensible. Plugins register commands with names, aliases, categories, and subcommands. Type a slash to trigger the command menu, filter by prefix, and execute with Enter. Commands can run instantly, show modals, take over the status area, or request inline input.

Built-in commands cover everything from profile management (/profile, /llm) to permissions (/permissions, /security) to tmux control (/terminal, /tmux). Plugins can add their own commands that feel native to the system.

Local-First by Design

Both Kollabor products store data locally. Conversations live in JSONL files at ~/.kollabor/ (App) and ~/.kollabor-cli/ (CLI). Configuration is local. Projects are workspace-aware. Nothing gets sent to the cloud without your explicit knowledge and consent.

We support local LLMs through OpenAI-compatible APIs, so you can run Ollama or LM Studio and keep everything on your machine. When you do use cloud APIs, we implement proper rate limiting, retry logic, and streaming responses.

MIT License: Real Open Source

Kollabor is released under the MIT License. That means you can use it, modify it, distribute it, even sell products built on it—no copyleft restrictions, no attribution requirements beyond the license itself. We believe developer tools should be genuinely free.

Get Started

Install Kollab CLI with pip: pip install kollabor

Download Kowork from our GitHub releases for macOS, Windows, or Linux.

Both products are in active development. We're shipping features rapidly, responding to feedback, and building in public. Join us on GitHub, star the repos, and help us build AI collaboration tools that respect your workflow and your freedom.

This is just the beginning. Welcome to Kollabor.