Key Takeaways
- Branching conversation interface: fork any AI message to explore an alternative response without losing the original
- Nested folder organization for chats, plus the ability to pin important conversation branches
- Hybrid model access: local inference via a separately running Ollama instance, or cloud models (Claude, GPT-4, and others) via your own API keys
- Chats and settings are stored in a local SQLite file on your device, not on Hanoki's own servers
- No account required to use the app
- Currently available for macOS only; Windows and Linux builds are not confirmed as shipped
📍 In One Sentence
Hanoki is a macOS desktop chat app with a branching conversation interface that lets you fork AI replies into alternative paths, organize chats into nested folders, and pin important branches — a hybrid app that runs models locally through your own Ollama instance or uses cloud models like Claude and GPT-4 through your own API keys, storing chats locally in a SQLite file.
💬 In Plain Terms
Instead of one straight-line conversation, Hanoki lets you branch off from any AI reply to try a different question, then keeps both versions available; it can talk to AI running on your own computer through a separate tool called Ollama, or to AI in the cloud if you supply your own paid API key, and it keeps your chat history in a file only on your machine.
📌Note: This review is based on Hanoki's official site and public product information, not independent hands-on benchmarking. No speed, accuracy, or pricing figures are included because none were independently verified for this article.
Is Hanoki Actually Local AI?
Hanoki is not a local-only AI app — it is a hybrid chat client, and local inference only happens when you separately set up and run Ollama. By default, Hanoki is built to talk to cloud AI models such as Claude and GPT-4 through API keys you provide and pay for yourself. Local inference is available, but only if you point Hanoki at a running Ollama instance on your own machine; Hanoki does not bundle or install a local inference engine itself.
- Cloud mode: connect your own Anthropic, OpenAI, or other provider API key; requests and billing go through that provider, not Hanoki
- Local mode: requires Ollama installed and running separately on your machine — Hanoki then sends requests to your local Ollama instance instead of a cloud API
- Hanoki itself does not run model inference — it is a chat client that either calls a cloud API or forwards requests to your local Ollama instance
- This differs from local-only apps like Atomic Chat, LM Studio, and Jan AI, which bundle their own local inference engine and run entirely offline by default without any separate tool
📌Note: Do not assume Hanoki keeps your conversation content off cloud servers by default. Unless you have set up and pointed Hanoki at a running Ollama instance, your messages are sent to whichever cloud provider's API key you configured.
What Is Hanoki?
Hanoki is a desktop chat application for macOS built around a branching conversation interface, designed to work with any AI model rather than being tied to a single provider. Published at hanoki.app, it lets you fork a conversation at any point to try an alternative prompt or response, then keeps both branches accessible instead of overwriting the original thread.
- Branching/forking: create alternative response paths from any point in a conversation instead of only editing and resending a message
- Nested folders: organize chats hierarchically rather than in one flat list
- Pinning: mark specific conversation branches as important so they stay easy to find
- Model-agnostic by design: works with cloud models via your own API keys, or local models via Ollama, rather than shipping with one built-in model
What Features Does Hanoki Include?
Hanoki's feature set centers on three things: forking conversations, organizing them into folders, and pinning the branches worth keeping track of. Each is aimed at making a branching chat tree manageable instead of turning into an unsearchable mess.
Where Does Hanoki Store Your Chats?
Hanoki stores chats and settings in a SQLite file on your own device, not on company servers, and does not require an account to use the app. This is a local-storage design rather than a local-inference design — it describes where your conversation history lives, not where the AI processing happens.
- Chat history and app settings are saved to a local SQLite database file on your machine
- No account or sign-up is required to use Hanoki
- Local storage applies regardless of whether a given chat used cloud models or a local Ollama model
Which Platforms Does Hanoki Support?
Hanoki currently ships for macOS only. The developers state that the underlying codebase supports Windows and Linux builds, but as of this review those builds are not confirmed as publicly shipped — do not plan around Windows or Linux availability today.
- macOS: shipped and available today
- Windows: not confirmed shipped — developers state the codebase supports it, but no public release has been verified for this review
- Linux: not confirmed shipped — same caveat as Windows
How Do You Set Up Hanoki?
Getting Hanoki running takes a handful of steps, and the path differs depending on whether you want cloud models, local models, or both.
- 1Download Hanoki for macOS from hanoki.app and install it.
- 2Open the app and decide whether you want cloud models, local models via Ollama, or both.
- 3For cloud models: obtain an API key from the provider you want to use (for example Anthropic for Claude, or OpenAI for GPT-4) and add it in Hanoki's settings — usage is billed to you directly by that provider.
- 4For local models: install Ollama separately, start it, and download a model through Ollama before pointing Hanoki at your running Ollama instance.
- 5Start a conversation. Use the fork/branch feature on any message to explore an alternative response, organize chats into folders, and pin the branches you want to keep easy to find.
Do I need Ollama to use Hanoki?
No — Ollama is only required if you want to run models locally. Without it, Hanoki works with cloud models through your own API keys.
Does Hanoki charge for API usage?
Hanoki itself is not documented as metering cloud API usage — you supply your own API key, and the model provider (for example Anthropic or OpenAI) bills you directly for that usage.
Who Should Use Hanoki?
Hanoki fits readers who specifically want a branching conversation structure on macOS, whether their models run in the cloud, locally via Ollama, or both.
Hanoki vs. Alternatives
Hanoki's closest comparisons are other easy, desktop-first chat apps aimed at non-technical users — but Hanoki's hybrid, branching-first design sets it apart from apps that bundle their own local inference engine.
App | Platforms | Local inference | Cloud models | Branching/forking |
|---|---|---|---|---|
| Hanoki | macOS only | Via separate Ollama install | Yes, via your own API keys | Yes — core feature |
| Atomic Chat | Mac/Win/Linux/iOS/Android | Built-in | No | No |
| LM Studio | Mac/Win/Linux | Built-in | No | No |
| Jan AI | Mac/Win/Linux | Built-in | No | No |
The key structural difference is that Hanoki is a hybrid, branching-first chat client rather than a bundled local-inference app: Atomic Chat, LM Studio, and Jan AI all include their own local inference engine and run entirely offline by default, while Hanoki requires a separate Ollama install for local inference and defaults to cloud models via your own API keys. This article has not independently benchmarked response speed or output quality across these apps and does not claim one is technically superior to another — the comparison covers documented platform, model-access, and feature facts only.
What Are Hanoki's Limitations?
Hanoki's branching interface is its clearest differentiator, but this review found several gaps worth naming directly rather than glossing over.
- No independent benchmarks of response speed, accuracy, or resource usage — none were measured for this article
- Not a one-click local AI app: local inference requires separately installing, running, and maintaining Ollama
- Cloud model use requires your own paid API keys from providers like Anthropic or OpenAI — Hanoki does not include free or bundled cloud usage
- macOS only today; Windows and Linux support is stated by developers as codebase-supported but not confirmed shipped
- No independently verified pricing information for this review — check hanoki.app directly for current terms
- No independent security audit of local SQLite storage or the app's network behavior was available for this review
Common Mistakes When Trying Hanoki
Most confusion around Hanoki comes from assuming it behaves like a local-only app, rather than from a flaw in the app itself.
Frequently Asked Questions
What is Hanoki?
Hanoki is a macOS desktop chat app built around a branching conversation interface — you can fork any AI reply into an alternative path, organize chats into nested folders, and pin the branches that matter.
Is Hanoki free?
This review does not have independently verified current pricing for Hanoki. Check hanoki.app directly for up-to-date pricing before deciding.
Is Hanoki a local AI app?
Hanoki is a hybrid app, not a local-only one. It defaults to cloud models via your own API keys; local inference is available only if you separately install and run Ollama.
Does Hanoki work with Ollama?
Yes. Point Hanoki at a running Ollama instance on your machine to use local models instead of, or alongside, cloud models.
What cloud models does Hanoki support?
Hanoki is designed to work with any cloud model you can access via your own API key, including models like Claude (Anthropic) and GPT-4 (OpenAI).
Where are my Hanoki chats stored?
Chats and settings are stored in a SQLite file on your own device, not on Hanoki's servers.
Does Hanoki require an account?
No. Hanoki does not require an account or sign-up to use the app.
What is message forking or branching in Hanoki?
Forking lets you create an alternative response path from any point in a conversation, so you can explore a different question or approach without losing the original thread.
Is Hanoki available for Windows or Linux?
Not confirmed as of this review. Hanoki currently ships for macOS only; developers state the codebase supports Windows and Linux, but no public build has been verified.
How is Hanoki different from Atomic Chat or LM Studio?
Atomic Chat and LM Studio bundle their own local inference engine and run fully offline by default. Hanoki is hybrid — it defaults to cloud models via your own API keys and needs a separate Ollama install for local inference — but adds a branching conversation interface those apps do not have.