Compare AI.
Explore ideas.

A Mac app for chatting with AI models, comparing their answers, and working through ideas together. Use an online AI service or run models on your own Mac.

Download for Mac How to get started

macOS 14 or later. No Xcode needed for the download. You’ll connect an AI service or install a local model before chatting.

What you can do

Compare answers

Ask 2–4 models the same question and read their answers side by side. Compare response times, usage, and estimated costs.

Work through an idea

Give AI participants different roles in a brainstorm. Shortlist ideas, challenge assumptions, and save a decision summary you can export as a PDF or Markdown text file.

Save an assistant

Create a “Ship,” a reusable assistant with a model, instructions, and reference material. Try a writing coach that remembers your preferred style.

Try another direction

Branch a conversation from an earlier message to explore a different answer without losing the original chat. Save prompts you want to use again.

Organize chats, Ships, and brainstorms into Personal, Work, or other Harbors. Connections and settings stay shared. The Getting Started tour guides your first conversation and can be resumed from Help.

More ways to customize

Choose a brainstorming method such as Six Thinking Hats or assign your own roles. Inspect the instructions and generation settings sent to a model, reuse prompt templates, and choose a light or dark theme.

Available settings vary by provider. Usage comes from provider reports where available; estimated counts are labeled and missing counts stay unknown. Costs are estimates, not billing totals; unavailable prices are labeled.

Your first conversation

Install ChatHarbor. Open the latest published release, download its DMG file, open it, and copy ChatHarbor to Applications.
Connect a model. Open ChatHarbor’s Settings → Models. For an online service, enter an API key from OpenAI, Anthropic, or Google Gemini and select Connect. An API key is an access code from the provider; usage may be billed separately from a chat subscription. For local models, follow the guide below.
Choose a model and start a chat. Select a connected model and open a new chat. Try: “Help me plan a small project. Ask me what you need to know first.”

Want to start without a cloud API key? Set up a local model.

Run AI on your Mac

ChatHarbor is the chat interface. Ollama or LM Studio runs the model. Install one of them and download a model that fits your Mac’s memory and storage. ChatHarbor does not include model files.

With Ollama

Install Ollama and keep it running. Download a model intended to run locally; avoid cloud models if you want to stay on your Mac.
In ChatHarbor, open Settings → Models → Ollama. Use http://localhost:11434 unless you changed Ollama’s address, then select Test.
Choose a downloaded Ollama model from the model list and start a chat.

Read Ollama’s setup guide

With LM Studio

Install LM Studio, download a model, and load it. In its Developer tab, start the local server.
In ChatHarbor, open Settings → Models → LM Studio. Check the prefilled name and server address, usually http://localhost:1234/v1.
Leave the key blank unless you enabled server authentication. Select Add, then choose an LM Studio model for your chat.

Read LM Studio’s server guide

Connected, but no models appear?

Check that Ollama or LM Studio is still running and that a model has finished downloading. For LM Studio, load the model and start its server. Match ChatHarbor’s address and authentication settings to the server, then reconnect.

Brainstorm locally, too

Open Brainstorm, choose a method, and assign a downloaded local model to every active role. You can use different models or give the same model several roles.

Use a local model for follow-up questions as well. Once the software and models are downloaded, local conversations can work offline. Cloud models and servers on other machines send content outside your Mac.

Loading several models can use more memory. Start with one model across multiple roles if your Mac struggles.

Where your data goes

Conversations are saved on your Mac. Provider API keys are stored in macOS Keychain.

Your messages and included reference material go to the model provider or server you choose. For on-device processing, choose downloaded models served from this Mac. Fetching a web page as reference material still requires an internet connection.

In-app update checks contact GitHub. You can disable automatic checks in Settings → General → Updates. Learn about updates or deleting local data.

Read the privacy policy

New in version 2.3.4

The latest release includes:

The auto-title setting now works in chats and Ship conversations.
Manually chosen conversation titles stay unchanged.
Ship history shows saved titles consistently.

Already on v2.3.2 or later? Choose ChatHarbor → Check for Updates… to get this version. You approve each installation and relaunch.

Using version 2.3.1 or earlier? Download this update manually once. After that, future versions can be offered inside ChatHarbor.

Download version 2.3.4 or read the changelog.

Can I use Apple Intelligence?

Apple Intelligence is available in source builds made with Xcode 26 or later, on compatible Apple Silicon Macs with macOS 26 or later and Apple Intelligence enabled. It is not included in the current published download.