AI Inspiration WorldWhat it is
rec is an inference engine from Recursive (GitHub organisation recursive-oss) for running Recursive's language models on your own computer. Once the command-line tool is installed, you can launch the pi or OpenCode coding agents directly, or use the serve command to offer an OpenAI-compatible API on your machine. By default it runs the Flint model locally, binds the API only to 127.0.0.1, and keeps sessions and settings on your machine. What sets it apart is that the engine, model and coding agents come set up together, so you can start coding locally right after install.
Who it's for:Developers with a high-end Apple Silicon Mac who want to code with a model running locally. If you don't write code, you won't need it.
Before you start: The official hardware requirement is a Mac with an M4 Pro, M5 Pro or M5 Max, at least 48GB of unified memory and about 30GB of disk space, so ordinary laptops and Windows won't work. pi also needs Node 22 and npm, and the docs and CLI are in English. The first run downloads the engine, model and Harness over the internet, with the model hosted on Hugging Face. The official install pipes a remote script straight into the shell, which runs code and changes your PATH. AI Inspiration World doesn't recommend running it unread: download install.sh and read it before installing. The model service keeps using memory in the background after you close the terminal, so stop it with the stop command. pi and OpenCode can go online and read and write files, so not everything stays on your machine. The engine code is under the Apache-2.0 licence; check the licences for the model weights and other components separately.
Based on the project's public documentation. Not tested by AI Inspiration World.