Port42 — AI Companions That Build on Your Mac
A native Mac app where you and multiple AI companions share live surfaces. Companions build interactive dashboards, tools, and apps mid-conversation that persist in your shared space.
TL;DR
TL;DR: Port42 is a native macOS app where you and multiple AI companions share live interactive surfaces — companions build dashboards, tools, and apps mid-conversation that appear instantly in your shared workspace.
Source and Accuracy Notes
⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.
- Project page: port42.ai — MUST visit and verify
- Source repository: none published (app is closed-source with open protocol)
- HN launch thread: news.ycombinator.com/item?id=47395572 (v0.5.0 launch)
- License: proprietary (local-first, end-to-end encrypted)
- Source last checked: 2026-08-06
What Is Port42?
Port42 is a native macOS application (macOS 15+, Apple Silicon) built around the concept of a port — a shared live surface where humans and AI companions work together in real time.
Unlike a chat interface where an AI responds and the conversation ends, Port42 companions build things during the conversation. A companion can create an interactive HTML/CSS/JS surface mid-conversation — a dashboard, a file browser, a monitoring tool — that renders inline in your shared space, pops out into a floating window, docks, resizes, and persists across sessions.
From the official site:
A native Mac app where you, your friends, and your AI companions share the same space. Talk, build, and watch ideas take shape together.
Companions can also act directly on your machine through the same API — clipboard access, screen capture, terminal execution — without switching contexts.
Supported Models and Integrations
Port42 is bring your own AI (BYOAI). It integrates with:
- Claude (Anthropic)
- Gemini (Google)
- OpenAI-compatible models
- LangChain agents
- OpenClaw
- Any Python framework
The app is also open by design — plugging in a new companion takes minutes. You can use your own memory, knowledge base, or agent runtime without lock-in.
Key Concepts
Ports — Shared Live Surfaces
A port is a live collaborative workspace. A companion builds whatever surface your task needs:
- A dashboard for tracking metrics
- A review room for code or documents
- A data explorer for querying and visualizing data
- A launch checklist with live progress
Surfaces render as live HTML/CSS/JS, created by the companion during the conversation and immediately usable.
Companions — AI Agents with Agency
A companion is not just an LLM. It can be:
- A language model
- A command or script
- A CLI tool
- A SaaS API integration
- A LangChain agent
- OpenClaw or any Python framework
Companions work alongside you — running your terminal, pulling data, building what you asked for. Multiple companions can collaborate on the same task, splitting work like a small team.
Action Surface — System-Level Acting
Companions can act on your machine directly:
- Read and write the clipboard
- Capture screenshots
- Execute terminal commands
- Access files and directories
This is done through a single API that works alongside the visual surface, so a companion can both show you something and do something without you switching apps.
Privacy and Architecture
Port42 is local-first and end-to-end encrypted. All data stays on your machine. The protocol is open, so you can plug in your own memory, knowledge store, or agent runtime. There is no data harvesting and no vendor lock-in — take your work with you when you leave.
Practical Evaluation Checklist
Getting started:
- [ ] Download the signed DMG from port42.ai (macOS 15+, Apple Silicon required)
- [ ] Launch the app and create your first port (shared live surface)
- [ ] Add a companion by connecting your own API key (Claude, Gemini, OpenAI-compatible)
- [ ] Ask the companion to build a surface — a dashboard, a tool, a checklist
- [ ] Watch the surface appear live in your shared space
- [ ] Pin, resize, or float the surface as needed
- [ ] Add a second companion to collaborate on the same task
Advanced:
- [ ] Connect a LangChain agent as a companion
- [ ] Use terminal companions that run scripts on your machine
- [ ] Set up a multi-companion review session (e.g. Maker + Critic pattern)
- [ ] Persist surfaces across sessions and organize your workspace
FAQ
Q: Is Port42 open source? A: The app itself is not open source, but the protocol is open. You can plug in your own memory, knowledge base, or agent runtime. There is no lock-in.
Q: Which AI models does it support? A: It supports Claude, Gemini, OpenAI-compatible models, LangChain agents, OpenClaw, and any Python framework. You bring your own API keys.
Q: Does it work offline? A: Yes, it is local-first. Your data stays on your machine and is end-to-end encrypted. The app works without sending data to external servers beyond your own AI API calls.
Q: What macOS version is required? A: macOS 15 or later, with Apple Silicon (M1/M2/M3/M4).
Q: Can multiple humans work with the same companion? A: Yes. Port42 is designed as a shared space — multiple humans and multiple AI companions can collaborate in the same port simultaneously.
Conclusion
Port42 reframes the AI assistant from a responder into a collaborator. Instead of copy-pasting AI output into your work, you work directly in a shared space where the companion builds the surfaces you need — a dashboard, a tool, a document — as the conversation unfolds. It is a genuinely different interaction model for AI on the desktop, and for Mac users who want local-first AI with real agency, it is worth trying.
For runany.dev readers interested in local AI setups, Port42 fits a specific niche: it is not a chatbot wrapper, it is a collaborative workspace powered by AI companions that build and act on your behalf.
Related Posts
ai-setup
Recall – Persistent Memory for Claude Code via MCP Hooks
Recall gives Claude Code a permanent memory store that survives session restarts and context compaction. Four hooks capture and restore context automatically — with cloud SaaS or self-hosted options.
2/28/2026
ai-setup
Sentrial – Catch AI Agent Failures Before Your Users Do
YC W26-backed AI agent observability platform. Trace sessions, detect silent regressions, and A/B test prompts in production before failures reach users.
5/28/2026
ai-setup
IonRouter – Fast Low-Cost AI Inference API
IonRouter is a YC W26 inference API routing open-source and fine-tuned models via an OpenAI-compatible endpoint, built on a C++ runtime optimized for GH200.
5/28/2026