Agent OS
Local-first AI team for building and shipping real software.
7 followers
Local-first AI team for building and shipping real software.
7 followers
Agent OS is an open-source, local-first AI project operating system. It coordinates specialized agents to plan, build, verify, recover, and ship real software—with persistent memory, browser evidence, and real deployment integrations.
Local-first with actual deployment hooks is a nice change from the usual cloud-only agent demos. Curious how the persistent memory holds up once a project gets messy.
@cansutokmalosf Thanks! I also think personal agents should run locally, while cloud platforms serve as external connectors.
The memory engine is what coordinates memory system and real project state. After each meaningful chat or task, the engine judges whether something deserves to be persisted into memory, then applies structured, concise updates to relevant memory files (based on run artifacts and other real evidence). So even when the project gets messy the memory usually won't get too messy :)
Would love a built-in dashboard showing which agents are working, what they've completed, and where they got stuck in real time, so I can step in without digging through logs. That kind of visibility would make complex multi-agent runs way easier to trust and debug.
@ferhat160021 Good advice. Visibility is truly important for multi-agent systems, especially when managing the status of different agents at the same time. Will add this feature to future versions!
Spent an hour with Agent OS and the persistent memory across sessions is genuinely useful, not just a gimmick. Surprised how well the browser evidence step caught a bug my own code review missed.
@muratftfl Wow thanks for having a try! Glad to hear that the memory system and browser verification are truly useful for real software projects. These are the core features I'm putting effort in, to give the system more long-term reliability.