Support & Privacy Policy

Last updated: September 17, 2026 · Effective date: September 17, 2026

Overview

LocalLLMClient is a local-first LLM chat app. Model inference runs on your device, so your prompts, conversations, and outputs are never sent to us.

For the complete policy, including exact storage paths and the full list of network requests, see the Privacy Policy.

Support

If something is not working, or you have a question, feature request, or privacy concern, contact us and we will help:

Email: vegaodm@outlook.com

When reporting a problem, including the model you selected and your iPhone, iPad, or Mac model helps a lot.

Frequently Asked Questions

Does the app work without an internet connection?
Yes. After a model finishes downloading, chat and tool calling work fully offline. Model files are downloaded once per model.
Is anything I type sent to a server?
No. Inference happens on your device. No prompt, message, attached image, or model output is transmitted anywhere.
What does use the network?
Only downloading model files. Before a download the app requests the repository's file list and file sizes so it can show a download size; then it downloads the weight files. Nothing else.
Which host downloads model files from?
The one you pick in the HF Source menu in the toolbar. The default is hf-mirror.com (a third-party mirror); you can switch to the official huggingface.co. Whichever you choose receives your IP address and standard request metadata.
Do I need an account?
No. There is no sign-in, and the app sends no credentials or access tokens.
Where is my conversation history kept?
On your device, in the app's container under Application Support/LocalLLMClientExample/History/, one file per model, text only. Messages with an attached image are not saved.
How do I delete my data?
Use Clear Chat to erase the current conversation, remove a model's data to drop its weights and history, or uninstall the app to remove everything.
Can I export a conversation?
Yes. The toolbar's Export… saves the current conversation as a Markdown file wherever you choose.

Data Collection

We do not collect any personal data.

  • All model inference runs entirely on your device
  • Prompts, conversations, attached images, and model outputs never leave your device
  • No analytics, telemetry, or crash reporting is sent to any server
  • No accounts, authentication, or tracking
  • No advertising or third-party data-broker SDKs

Network Access and Data Recipients

The app contacts a model host only to download model files. The host you select is an independent party to us and acts as its own data controller for those requests.

SelectionHost contactedWho operates it
Default hf-mirror.com A third-party mirror operator — not Hugging Face.
Optional huggingface.co Hugging Face, the model hub operator.

That host receives your IP address, standard request metadata such as the User-Agent string, and the public repository and file names being downloaded. It receives none of your conversation content.

Compliance note: the operator of the host you select — by default the operator of hf-mirror.com, an independently operated third-party mirror — processes your download request as a separate data controller. If you switch the source to huggingface.co, that party is Hugging Face. We neither control nor store any of this data and have no access to it.

Local Storage

  • Conversation history — stored locally in the app container at Application Support/LocalLLMClientExample/History/; text messages only
  • Model weights — stored at Documents/.localllmclient/huggingface/models/ on iPhone and iPad, and ~/.localllmclient/ on Mac
  • Diagnostics log — Documents/diagnostics.log (512 KB cap) holding timestamps and memory usage only, never prompt text
  • You can delete all data via Clear Chat, by removing a model's data, or by removing the app
  • Conversation history is retained until you delete it or uninstall the app
Device backups are separate from this app. If iCloud or Finder/iTunes device backup is enabled, your operating system may include app data in that backup. That transfer is performed by Apple under your Apple account and backup settings, and is governed by Apple's privacy policy — not by this app.

Third-Party Components

This app uses:

  • llama.cpp (C++ inference engine)
  • MLX (Apple's machine learning framework)
  • FoundationModels (Apple's on-device LLM framework, iOS 26 / macOS 26 and later)

These components run locally and do not transmit your conversations or send any data to their maintainers.

Your Rights

Since we do not collect data, there is nothing on our end for you to access, correct, or delete. All of your data stays on your device, and you can remove it at any time.

Children's Privacy

This app is intended for general audiences and does not knowingly collect personal information from children.

Policy Updates

If we make changes to this policy, we will update the "Last updated" date at the top of this page and, where appropriate, notify you in the app.

Contact

If you have questions or concerns about this privacy policy, please contact us at:

Email: vegaodm@outlook.com