Local AI Server
Built for developers

Your Mac.
Your AI.

Bring Apple Intelligence to your apps through a local API powered by Apple’s on-device Foundation Models.

Prompts, images, and responses are processed on your Mac. What you share stays between you and your Mac.

Try the local chat Coming to the Mac App Store Explore the app

Apple silicon · macOS 26+ · Apple Intelligence enabled

Running on this Mac
Local AI Server Chat showing a response to a request for hiking app names.
01 Chat with the on-device model

Build with on-device AI

Local intelligence for your workflows.

Build text-processing features with the Apple Intelligence model managed by macOS. Local AI Server keeps inference on your Mac and makes it available to your apps through a loopback HTTP API.

01 / Text

Translate and transform.

Translate between supported languages, rewrite drafts, or adapt the tone of a message. Send text from your app to the local model through the API.

02 / Workflows

Summarize and categorize.

Turn notes into summaries, classify incoming text, or identify key information in short documents. Prompts and responses are processed on the Mac.

03 / Vision

Ask about an image.

On macOS 27 or later, attach an image when the system model reports vision support. Image analysis runs on your Mac and returns a text response.

Context depends on the system model. Supported configurations include 4,096- and 8,192-token windows. Diagnostics shows the exact limit reported by your Mac. Instructions, conversation history, and generated responses share that budget.

For developers

Apple Intelligence on localhost.

Send requests to Apple’s on-device model through a subset of OpenAI-style Chat Completions, including streaming text responses. Use the default API address or set your own port.

Copy examples for JavaScript, Node.js, Python, C#, cURL, Swift, and Ruby from Docs. Use Diagnostics to check the model’s context size and capabilities, and the native chat to test prompts with the API switched off.

When the service is on, apps and browser pages that can reach loopback can use it without an API key. Your browser may ask for local network access. Turn the service off when you’re done.

Terminal
curl -sS http://127.0.0.1:8765/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "apple-on-device",
    "messages": [
      {"role": "user", "content": "Suggest a name for a hiking app."}
    ]
  }'
Service panel showing the on switch, the local API address, and model status.
ServiceStart, stop, and configure your local API.
Docs panel showing a language picker and a copyable API example.
DocsCopy examples right from the menu bar.

Support

Need a hand?

Local AI Server is made by Daniel Atik. For help, include your macOS version and describe what happened.

Email support
What do I need to run it?

A Mac with Apple silicon, macOS 26 or later, and Apple Intelligence enabled and ready. Image input requires macOS 27 or later and a system model that reports vision support.

Do I need to download a separate model?

Local AI Server uses the on-device model supplied by macOS. If Apple Intelligence is still being set up, macOS handles the initial model download.

Why can’t my app reach the API?

Open Local AI Server from the menu bar, turn on Service, and copy the address shown there. The API listens on the same Mac. Browser pages may need permission to access loopback.

How do I delete a conversation?

Open Chat, choose the conversation, and delete it from the Conversations menu. See the Privacy Policy for details.

About Local AI Server

Developed by Daniel Atik

Local AI Server demo