Skip to content
Get packs

Guide · Open source

How to use Strata: install, launch and connect local AI

Updated · written with AI from the project's own documentation and checked against our catalogue data

Strata is a free, MIT-licensed app and terminal tool for running large language models on your own computer. This guide explains how to install it, start a model, and connect coding tools such as Claude Code or Codex CLI to the local model. By the end you will know which install path to pick and how to check that your hardware is supported.

StrataNiko1221/Strata · ★ 38.1k · MIT

Quick start

  1. 1Check your hardware. Confirm your GPU or Mac is on the supported hardware list and that you have enough RAM and free disk space for the model you want.
  2. 2Install Strata. Download the desktop app from the releases page, or run the install command for Linux, Mac or Windows PowerShell.
  3. 3Open a new terminal (Windows). If you used the Windows terminal install, open a new terminal window before running Strata commands.
  4. 4Launch Strata. Run strata launch and follow the prompts to download and start a model.
  5. 5Chat or connect a tool. Use the chat in the app, or run strata connect to set up a coding tool against the local API.

What Strata is and who it suits

Strata runs AI models locally, so chat, coding and image work happen on your own hardware and no account is needed. It exposes OpenAI- and Anthropic-compatible APIs on localhost, which lets other tools use the local model.

It suits people who want private AI chat, developers who want coding tools to talk to a local model, and users with a supported NVIDIA RTX or Apple Silicon Mac who want a one-click installer. It is less suited to older or low-memory machines, because large models need a lot of RAM or disk space.

Check that your hardware is supported

Before downloading a large model, check the hardware list. The README names specific supported GPUs, including NVIDIA GeForce RTX 20, 30, 40 and 50 series, several AMD Radeon cards, Intel Arc on Linux, Apple Silicon M1 or newer, AMD Ryzen AI Max (Strix Halo) and NVIDIA DGX Spark on Linux.

  • NVIDIA GeForce RTX 20 / 30 / 40 / 50 series
  • AMD Radeon RX 7700 / 7800 / 7900, RX 9060 / 9070, AI PRO R9700
  • AMD Radeon RX 6800 / 6900 (experimental)
  • Intel Arc (Linux)
  • Apple Silicon, M1 or newer
  • AMD Ryzen AI Max (Strix Halo)
  • NVIDIA DGX Spark (Linux)

Install Strata with the desktop app or the terminal

The desktop app is the simplest route. Download the installer for Windows, Linux or Mac from the releases page, install it, choose a model and start chatting. On first start the app downloads the engine for your graphics card.

For the terminal, run the install command for your system. On Windows, open a new terminal after installing.

  • Desktop app: download from the releases page for Windows, Linux or Mac
  • Linux / Mac: run the curl install script shown below
  • Windows: run the PowerShell install command shown below
curl -fsSL https://github.com/Niko1221/Strata/releases/latest/download/install.sh -o install.sh && sh install.sh

Start a model in the terminal

After installing, run the launch command. Follow the prompts to download and start a model. The README's terminal guide covers the details.

For the DeepSeek model, the README shows a launch command with explicit family and model flags and a plan option. Model downloads can be very large, so check the disk space figures first.

strata launch

Connect coding tools to the local model

Strata's API is compatible with OpenAI and Anthropic clients, so tools such as Claude Code, Codex CLI, OpenCode and others can use the local model. Run the connect command to see setup instructions for your tool.

The connection guide in the README has the specific steps for each tool.

  • Claude Code
  • Codex CLI
  • OpenCode
  • oh-my-pi, pi.dev, Hermes Agent and OpenClaw
strata connect

Know the speed trade-offs

Speed depends heavily on hardware and model. The README reports output speeds of about 4 to 25 tokens per second for DeepSeek across several setups, with prompt reading faster. Lower-memory models such as the Gyro packs use less memory but run more slowly.

Measured figures for each setup are listed in the README; treat them as examples rather than guarantees.

What people use it for

  • Chatting with an AI model privately on your own computer
  • Letting Claude Code or Codex CLI use a local model through the localhost API
  • Working with images when a model supports image input
  • Trying large open models without a cloud subscription

Frequently asked questions

Is Strata free?

Yes. The README describes Strata as free and open source under the MIT licence, and says no account is needed. Model packs come from third-party providers and each has its own licence, listed in the credits docs.

What hardware do I need to run Strata?

The README lists specific supported GPUs, including several NVIDIA RTX, AMD Radeon, Intel Arc and Apple Silicon options. Large models need plenty of RAM or disk space, so older or low-memory machines may not be a good fit.

How do I install Strata on Linux or Mac?

Run the curl install command from the README in your terminal, then run strata launch. The README also offers a desktop app installer for each system.

How do I install Strata on Windows?

Run the PowerShell command irm with install.ps1 from the README, open a new terminal, then run strata launch.

Can Claude Code or Codex CLI use Strata?

The README says Strata exposes OpenAI- and Anthropic-compatible APIs on localhost, and that tools such as Claude Code and Codex CLI can use them. Run strata connect for setup instructions.

How does Strata compare to Ollama or LM Studio?

Strata is a local model runner like Ollama, LM Studio, llama.cpp and GPT4All. The README does not compare them directly, so test each on your own hardware and pick the one whose model catalog and tools suit you.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Strata suits people who want a free, private local AI app on a supported NVIDIA RTX or Apple Silicon machine, or developers who want coding tools to use a local model. It is a poor choice if your computer is older or low on memory, if you need the fastest answers, or if you need guaranteed support and a clearly documented model catalog.

Install options & downloads