Strata
Runs a local AI chat and coding model on your own computer, with an OpenAI/Anthropic-style API on localhost.
- Stars
- 38.1k
- Forks
- 4.1k
- Licence
- MIT
- Language
- C++
- Last update
- 6 days ago
- Type
- App
What it is
Strata is a free, MIT-licensed app and terminal tool that runs large language models locally on your own hardware, so you can chat, code, and work with images without an account. It exposes OpenAI- and Anthropic-compatible APIs on localhost so other tools such as Claude Code or Codex CLI can use the local model. The README says it targets a range of NVIDIA, AMD, Intel, and Apple Silicon hardware.
- Chat with a locally running model in the desktop app
- Point a coding agent such as Claude Code or Codex CLI at a local OpenAI-compatible endpoint
- Send images to a model that supports image input
- Run the terminal app with strata launch and monitor speed and memory
Great for
- People who want to run AI chat privately on their own computer
- Developers who want coding tools to talk to a local model
- Users with a supported NVIDIA RTX or Apple Silicon Mac who want a one-click installer
- Anyone wanting to try large open models without a cloud subscription
Not ideal for
- People with low-memory or older machines, since the README lists specific supported GPUs and large models need lots of RAM or disk space
- Users who need the fastest possible answers; the README reports only 4 to 25 tokens/s for DeepSeek on several setups
- Users who need guaranteed support or a clearly documented model catalog, since the README links to many model packs and details live in separate docs
How to get it
- 1
Download the app
Linux / Mac terminal install
curl -fsSL https://github.com/Niko1221/Strata/releases/latest/download/install.sh -o install.sh && sh install.shWindows PowerShell terminal install
irm https://github.com/Niko1221/Strata/releases/latest/download/install.ps1 | iexStart the terminal app
strata launch - 2
Run it
- Download the installer for your system from the latest release, or use the terminal install command
- Install the app and choose a model
- Let the app download the engine for your graphics card on first start
- Start chatting in the desktop app, or run strata launch in the terminal
- 3
Get the most out of it
- Run strata connect to get setup instructions for connecting tools like Claude Code or Codex CLI
- Check docs/MODELS.md to pick a model size that fits your memory
- Use the terminal monitor to watch model activity, speed, and memory
- Read docs/TROUBLESHOOTING.md if the install or model start fails
Downloads
Strengths
- Free and open source under the MIT license, with no account required per the README
- Offers both a desktop app and a terminal workflow with installers for Windows, Linux, and Mac
- Exposes local OpenAI- and Anthropic-compatible APIs for existing tools
Trade-offs
- Hardware support is specific and many models need large memory or disk space, per the README
- Performance varies a lot by hardware, and some models are only documented for particular GPUs
- The README is truncated and the model landscape is complex, so setup may require reading several linked guides
How it compares
| Project | Stars | Type | Level | Updated | Best for |
|---|---|---|---|---|---|
| Strata (this) | 38.1k | App | beginner | 6 days ago | People who want to run AI chat privately on their own computer |
| ollama | 186k | App | beginner | 4 days ago | Developers who want a local model API for their own apps |
| llama.cpp | 133k | Library | intermediate | 4 days ago | Developers who want to run open-weight models locally without a cloud API |
| gpt4all | 77.3k | App | beginner | 1.4 years ago | People who want to chat with a language model without sending their text to a cloud service |
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.