Minutus
RUN YOUR OWN LOCAL AI

Your local
AI workspace.

Install Minutus, download a model, and run your own AI on your Mac.
Chat, understand images, and work with your files using local inference.

Build from source

32 GB unified memory recommendedApple Silicon · macOS 13.3+ · Models downloaded separately

Minutus desktop interface showing a project sidebar and a scripted sample chat about planning a morning.

Actual interface · scripted sample conversation

Local inference Permissioned tools x402 support Open source

EXAMPLES

What you can do
with Minutus.

Local work, with external tools
when you choose to use them.

01 / PROJECT FILES

Turn project notes
into a task list.

Give Minutus access to a project folder. Extract decisions, list open questions, or draft next steps using local inference.

“Read these meeting notes. List the decisions and action items.”
File access stays within your granted scope.
Project notesIllustrative workflow
meeting-notes.mdDecisions
and action items
Notes → task list
Selected workspace only
02 / SCREENSHOTS

Explain an error
from a screenshot.

Attach a screenshot and ask what it means. Get an explanation or a suggested next step with image understanding on your Mac using the matching vision module.

“What does this import error mean, and what should I try next?”
Image understanding runs locally.
Attached screenshotIllustrative workflow
import-error.pngAsk about the error and the next step.
Process the image on your Mac
03 / WEB RETRIEVAL

Check the docs.
Summarise the changes.

Retrieve public documentation when a task needs information beyond your local files. Review the destination before retrieval, then keep the source references with the answer.

“Read these release notes and list the changes that affect this project.”
Web features contact external services.
Public sourcesIllustrative workflow
Official documentationRetrieve the relevant pages
Release notesIdentify relevant changes
A summary with source references.
Sources remain available for review
04 / X402 RESOURCES

Use a paid endpoint.
Review the payment.

Use supported public x402 services as part of a task. Review the request, then separately approve the payment in a native dialog.

“Use this x402 endpoint for the task. Show me the payment request for review.”
USDC on Base mainnet.Compatibility details
x402 requestIllustrative workflow
service.example/report
Network
Base mainnet
Asset
USDC
Payment
Native review

Request → review → approve or cancel

Model output cannot approve payment

HOW IT WORKS

Local processing.
Explicit permissions.

Inference stays local.

Chat and image understanding run on your Mac. Optional dictation uses a separate local speech model.

File access is scoped.

You choose which workspace Minutus can access. The app enforces file permissions independently of model output. Folder access starts read-only.

x402 payments need approval.

Supported x402 requests use separate native request and payment review. Current compatibility is USDC on Base mainnet.

x402 compatibility

Model downloads, web tools, and other optional network features contact external services. How privacy works

MAC & RAM REQUIREMENTS

Will it run
on your Mac?

Apple Silicon · macOS 13.3+
RAM is separate from disk space.

32 GB unified memory recommended

For the current model catalogue, we recommend a Mac with 32 GB of RAM. The model and context length affect how much memory you need.

8 GB and 16 GB Macs are below the current catalogue’s base recommendations.

Current catalogue RAM guidance
ModelBase recommendation
Qwen3.8 27B
GGUF · recommended		30 GiB
Muse Glimmer 30B
GGUF		24 GiB

Vision and draft modules can add 2 GiB each where applicable. Longer context and other apps need additional headroom. These are recommendations, not a verified minimum.

Complete model bundles are about 16.5 GiB for Qwen and 18.4 GiB for Glimmer, including declared companions. Allow additional disk space for the app and runtimes. Experimental MLX is disabled in release builds.

GETTING STARTED

Run AI
on your own Mac.

01

Build MinutusGet the source on GitHub and follow the README.

02

Open your local buildLaunch the built app to begin setup.

03

Add a modelDownload a supported model in Minutus, or import matching local files.

MINUTUS ON GITHUB

Get Minutus.
Build it locally.

Get the source on GitHub and follow the README to build locally.

Build from source32 GB unified memory recommendedApple Silicon · macOS 13.3+ · Models installed separately
RAM requirements Source release