Pocket LLM StudioPREVIEW Get the app

BUILT FOR THE MACHINE YOU ALREADY OWN

Your local AI.
A little less local.

Chat with your model, watch the files change, and get hands-on with your code. All on your computer. Even when you aren't.

Download for Windows See how it works

Preview v0.1.0 · Windows x64 · Bring your own model

No Node.js or terminal needed to install.
Illustrative interface preview. Your model, your approved project folders.

Your model.LM Studio or a compatible API

Your real files.Not a detached code sandbox

Your workbench, anywhere.Desktop companion + mobile web app

MORE THAN A CHAT WINDOW

Keep the conversation.
Stay close to the code.

A useful local AI workflow needs more than a place to send prompts. Pocket LLM Studio brings the conversation and the project into the same view.

From prompt to actual project files.

Let a tool-capable model inspect and update files in the folders you approve. Follow its progress, then review and edit the result yourself.

File tools only. Arbitrary shell commands are not enabled.

Your model, your choice.

Choose among the models your server advertises. Switch for the next prompt without throwing away the conversation.

See that it's working.

Watch streamed responses, tool activity, and reasoning when the model exposes it. Stop a request or retry an earlier prompt.

An editor, not just an attachment.

Open multiple file tabs with independent drafts and undo history. Syntax highlighting and revision-checked saves help you review changes.

Built for a smaller screen.

Move between Chat, Files, and Editor on your phone. Use focus mode for a conversation with a little more breathing room.

A workspace for each project.

Register existing folders, choose a startup default, and let each connected device work in its own selected workspace.

START ON YOUR COMPUTER

A small setup.
A much closer workspace.

The Windows companion connects the model to your files. The mobile web app is your window back to that computer.

  1. 01

    Install the companion.

    Download the Windows x64 installer and open Pocket LLM Studio. It includes its runtime; you don't need Node.js or a terminal.

    Get the Windows preview ↓
  2. 02

    Bring a model and a folder.

    Run a tool-capable model in LM Studio. In companion Settings, enter its API URL and choose an existing project folder to share.

    http://127.0.0.1:1234/v1
  3. 03

    Optionally, take it with you.

    Configure a relay and its registration token, then use Pair devices to approve your phone. Keep the companion open and the computer awake.

    Read the remote-access requirements ↗
Before you start

64-bit Windows for the companion. A running, tool-capable model server. An existing project folder you intend to share. For mobile use, a modern browser and an active connection back to your computer.

YOUR COMPUTER STAYS IN THE LOOP

Local doesn't have to
mean out of reach.

Keep inference on your hardware when you use a local model. Choose which project folders are shared, approve remote devices, and review the files your assistant changes.

Understand the preview's boundaries ↗

Explicit workspace accessOnly register folders you intend connected devices and the model to read and edit.

Pairing, not an open file serverRemote sessions use approved devices and an encrypted relay connection.

You can still get hands-onReview the output in the editor. Revision checks flag stale saves rather than silently overwriting newer files.

MEET YOUR NEW LOCAL WORKBENCH

Your next idea.
Your own machine.

Start with the desktop companion. Connect your model, pick a project, and see what you can build.

WINDOWS x64 / EARLY PREVIEW
Pocket LLM StudioVersion 0.1.0PREVIEW
Download Windows installer

Download via GitHub Releases

✓ Self-contained installer✓ No Node.js required✓ Model runs separately

This build is unsigned. Install only releases you trust. This is an early preview, not a finished production release.

View GitHub release ↗

A FEW THINGS WORTH KNOWING

Before you
dive in.

Local models, real files, and a few practical requirements.

Does the download include an AI model?

No. The installer includes the companion's runtime, but LM Studio and the model are separate. Load a tool-capable chat model and point the companion at its OpenAI-compatible API URL. Model capabilities and response quality depend on the model you choose.

Learn about LM Studio ↗

What do I need to use it from my phone?

Your computer and model server must remain running, and your phone needs an active connection. Remote relay access currently requires a configured relay and its registration token, followed by QR pairing and computer-side approval. A private-network setup such as Tailscale is another option.

The prefilled relay address does not include a registration token. Downloading the companion does not automatically provision a managed relay account. The phone client is a web app/PWA, not a separate App Store download.

Can it run commands or tests on my computer?

No arbitrary shell or process execution is enabled. The model's tools can list, read, and write accessible project files. Use your normal development tools to run commands, builds, and tests, and review generated edits before using them.

What should I share with the model?

Only add folders you intend to expose to the assistant and approved devices. Use a disposable project or a backed-up working copy while evaluating the preview. File changes can be saved by the model without a separate approval step. Revision checks are not a substitute for backups or version control.

Choose your model API carefully: a local endpoint keeps inference on your own hardware, while a remote model provider receives the prompts and file context you send to it.

Is there a Mac or Linux installer?

Not currently. The downloadable companion is a Windows x64 build. The mobile client runs in a browser and can be installed as a PWA where the browser supports it.

Is this ready for production?

This is an early preview. The Windows installer is unsigned; signed releases, automatic updates, and an independent production security review have not shipped. Windows may flag an unsigned installer. Only install software from a source you trust, and do not disable operating-system security protections.

Published installers are distributed through GitHub Releases. Local website previews serve a verified build from the author's computer instead.