Introducing a recent labor of love to the world... Maestro is a cross-platform desktop app for orchestrating your fleet of Al agents. Set them loose on complex tasks, check in from your phone, and let them work while you sleep. Free and open source.
I strongly prefer interacting with ReAct (reason-act) agents over chat agents. It allows for file-system based memory, tool creation and use, MCP agents, etc. I have so many parallel threads with so many agents that I lose track of them regularly. This was the impetus behind the creation of Maestro. Now all my agents sit side-by-side, each logical thread in its own tab, and keyboard short cuts galore allow me to conduct them all at lighting speed.
The single most powerful feature of the application is the Auto Run capability. Work with Al to generate a series of detailed implementation plans, then execute on them with a fresh context per task, allowing for nonstop uninterrupted execution. The current record is over two days of runtime! Even more powerful, organize multiple Markdown documents into a loop-able Playbook, with one stage creating work for other stages.
Just released Group Chat capability in v0.10.0, allowing one to communicate with a team of agents in a single thread.
Mostly tested on OSX with Claude. Codex and Open Code support was recently added and is maturing though not as full fledged as Claude. Please download and send me feedback via issue, PR, Discord message, smoke signal, singing telegram, carrier pigeon, etc.
Introducing a recent labor of love to the world... Maestro is a cross-platform desktop app for orchestrating your fleet of Al agents. Set them loose on complex tasks, check in from your phone, and let them work while you sleep. Free and open source.
I strongly prefer interacting with ReAct (reason-act) agents over chat agents. It allows for file-system based memory, tool creation and use, MCP agents, etc. I have so many parallel threads with so many agents that I lose track of them regularly. This was the impetus behind the creation of Maestro. Now all my agents sit side-by-side, each logical thread in its own tab, and keyboard short cuts galore allow me to conduct them all at lighting speed.
The single most powerful feature of the application is the Auto Run capability. Work with Al to generate a series of detailed implementation plans, then execute on them with a fresh context per task, allowing for nonstop uninterrupted execution. The current record is over two days of runtime! Even more powerful, organize multiple Markdown documents into a loop-able Playbook, with one stage creating work for other stages.
Just released Group Chat capability in v0.10.0, allowing one to communicate with a team of agents in a single thread.
Mostly tested on OSX with Claude. Codex and Open Code support was recently added and is maturing though not as full fledged as Claude. Please download and send me feedback via issue, PR, Discord message, smoke signal, singing telegram, carrier pigeon, etc.
Cheers
-pedram