A macOS client for the pi coding agent.
It spawns pi --mode rpc as a sandboxed subprocess and renders the conversation
with native macOS frameworks.
Status: pre-alpha; already what yours truly uses for pi on a daily basis.
On Mac OS, follow the Pi Quickstart.
This project does not configure Pi or your authentication with an LLM provider.
Then, build and run with:
./run.sh # generate the project, build, quit any running instance, launch
On first launch the app walks you through setting up the agent's sandbox:
After that, pick a project and start prompting; the sandbox is applied uniformly to all sessions.
A settings page is available to change the sandbox policy, taking effect on next launch.

The agent runs inside a Seatbelt sandbox (default-deny): the policy defines everything it can touch, and everything else is off-limits — files, network, syscalls. Configuration lives in Settings (app menu → Settings…).
pi --mode rpc
(default: pi on PATH); point it at a different binary from Settings.100 commits
Swift
99.2%
A macOS client for the pi coding agent.
It spawns pi --mode rpc as a sandboxed subprocess and renders the conversation
with native macOS frameworks.
Status: pre-alpha; already what yours truly uses for pi on a daily basis.
On Mac OS, follow the Pi Quickstart.
This project does not configure Pi or your authentication with an LLM provider.
Then, build and run with:
./run.sh # generate the project, build, quit any running instance, launch
On first launch the app walks you through setting up the agent's sandbox:
After that, pick a project and start prompting; the sandbox is applied uniformly to all sessions.
A settings page is available to change the sandbox policy, taking effect on next launch.

The agent runs inside a Seatbelt sandbox (default-deny): the policy defines everything it can touch, and everything else is off-limits — files, network, syscalls. Configuration lives in Settings (app menu → Settings…).
pi --mode rpc
(default: pi on PATH); point it at a different binary from Settings.100 commits
Swift
99.2%