Troubleshooting

Solutions for the most common issues. If a problem isn't covered here, contact support and include your macOS version, Mac model, and the model you were running.

App won't open

"LocalChat can't be opened because Apple cannot check it for malicious software"

macOS Gatekeeper blocks apps downloaded from the internet until you approve them once:

  1. Open System Settings > Privacy & Security
  2. Scroll down to the message about LocalChat
  3. Click Open Anyway
  4. Enter your password if prompted

This only needs to be done once.

App opens then immediately closes

  1. Restart your Mac and try again
  2. Make sure you're on macOS 13 (Ventura) or later
  3. Delete the app and reinstall from the download page - your conversations and models are stored separately and survive a reinstall

Model downloads

Download fails or stalls

  1. Check your internet connection
  2. Check free disk space - models need their full size available, plus some headroom
  3. Cancel the download and start it again; progress resumes where it stopped
  4. If a specific model keeps failing, try a smaller one first to confirm downloads work at all

Download is slow

Model files are large (0.3-19 GB). Speed depends on your connection, not on LocalChat. Downloads continue in the background while you use the app with an already-loaded model.

Memory and performance

"Not enough memory" or model fails to load

Each model card lists the RAM it needs. If loading fails:

  1. Close memory-heavy apps (browsers with many tabs are the usual culprit)
  2. Try a smaller model - a 4B model needs roughly half the memory of a 9B one
  3. Lower the context size in Settings > Response Style - larger context windows use significantly more memory
  4. Restart LocalChat so memory from the previous model is fully released

Responses are slow

  • Check the model size - Bigger models are slower. For quick tasks, a small model is often the better trade
  • Check Speed mode - In Settings > Response Style, set it to Maximum Speed to use full GPU acceleration
  • Check what else is running - Video calls, exports, and builds compete for the same GPU and memory
  • Intel Macs - Response speed on Intel is much lower than on Apple Silicon; smaller models help

Mac gets hot or fans spin up

Running a model is heavy work, comparable to gaming or video rendering. It's normal under load. To reduce it, use a smaller model or switch Speed mode to Balanced or Power Saver.

Documents and files

The AI can't see my attached file

  1. Confirm the file type is supported: PDF, MD, CSV, XLSX, PNG, TXT
  2. Re-attach the file in the current conversation - files attach per chat unless they're in a project
  3. For scanned PDFs with no text layer, use a vision or OCR model so the pages can be read as images

Answers ignore parts of a long document

Long documents can exceed the model's context window. Raise the context size in Settings > Response Style, or add the document to a project so it's indexed and searched instead of read end-to-end.

Conversations

Chat history is missing

Conversations are stored locally. They don't sync between Macs, and reinstalling macOS or moving to a new machine without a backup will lose them. Use Settings > Data & Export to export conversations as JSON before migrating.

Model gives strange or repetitive output

  1. Start a fresh conversation - very long chats can degrade output
  2. In Settings > Response Style, raise Avoid repetition
  3. Try a different model; some handle certain tasks better than others

License

License key isn't accepted

  1. Check for typos - copy and paste the key from your purchase email
  2. Make sure you're not exceeding the device limit for your license
  3. Still stuck? Contact support with your purchase email address
  • Installation - Requirements and install steps
  • Models - Choosing a model that fits your Mac
  • Settings - Performance and memory options