I tried @pidotdev over the past few days and fell in love with its philosophy: a small core, extensible by design. So I built Pi Desktop, an open-source GUI powered by the real Pi runtime. It’s still alpha, and I’d love your feedback:
https://t.co/DuckYqjkB8
@il_thirtyseven@pidotdev Right now the most polished experience is on macOS (both Apple Silicon and Intel).
Windows and Linux builds are available too (check the latest release), but macOS is where I’ve spent most of the testing so far.
Would love to hear how it goes on your side!
I tried @pidotdev over the past few days and fell in love with its philosophy: a small core, extensible by design. So I built Pi Desktop, an open-source GUI powered by the real Pi runtime. It’s still alpha, and I’d love your feedback:
https://t.co/DuckYqjkB8
Thanks! We’re actually working on this. The goal is to let Windows users run Pi on a remote server through RPC over SSH, while keeping the desktop app as the local interface. Remote agents and projects are part of the direction too, with a workflow as smooth as Codex Desktop but still true to Pi’s lightweight, extensible philosophy.
Thanks! I’m focusing on a more native desktop workflow: better project and session management, embedded terminal and browser tools, easier provider and package setup, and deeper file and Git integration, while keeping Pi’s package-first philosophy. Just like Pi, you can also download the source and ask any LLM inside the app to add or modify features for you. What features do you use the most?
I tried @pidotdev over the past few days and fell in love with its philosophy: a small core, extensible by design. So I built Pi Desktop, an open-source GUI powered by the real Pi runtime. It’s still alpha, and I’d love your feedback:
https://t.co/DuckYqjkB8
I tried @pidotdev over the past few days and fell in love with its philosophy: a small core, extensible by design. So I built Pi Desktop, an open-source GUI powered by the real Pi runtime. It’s still alpha, and I’d love your feedback:
https://t.co/DuckYqjkB8
I’ve just published Tartarus, a local runtime for orchestrating multiple coding agents through MCP.
The idea comes from a simple problem: many developers already use tools such as Claude Code, Codex, Cursor, Pi, OpenCode, or other CLIs, but making them work together requires different configurations, process management, duplicated context, and often a significant amount of wasted tokens.
Tartarus does not try to replace the main agent, and it does not introduce another artificial “brain.”
The orchestrator remains Claude, Codex, or Cursor.
Tartarus handles the operational primitives instead:
• detects installed and already authenticated CLIs
• creates isolated Git worktrees
• prepares structured context packs
• starts one or more workers in parallel
• collects logs, status, diffs, and handoffs
• leaves the final decision to the main agent
That is the core idea:
Tartarus does not replace your coding agent. It gives it more hands.
The project is still evolving, especially around security, job recovery, and packaging, but the foundation is already working and completely open source.
Repository:
https://t.co/yVZB34BC5u
Feedback, criticism, and contributions are welcome.
#opensource #aiagents #codingagents #mcp #developerTools #typescript #softwareengineering #github