omnigent-ai/omnigent

Omnigent sits one layer above the individual AI coding agents most teams are already using and lets you treat them as interchangeable workers. The concrete value: when Anthropic ships a breaking change to Claude Code or you want to A/B Codex against a local model, you swap the harness config rather than rewriting your automation layer. Policy enforcement and sandboxing live at the orchestrator level, so you get a consistent audit trail regardless of which agent ran the actual task. The real-time collaboration angle means a solo founder and a contractor can share the same agent session without screen-sharing hacks. This is early and the repo will roughen before it smooths, but the architecture is solving a real problem — agent fragmentation is already painful for anyone running more than one coding assistant. Reservation: the project is very new and production readiness is an open question. Evaluate it as infrastructure you contribute to, not infrastructure you depend on today. -> Best for: AI engineer or SaaS team of 2-5 running multiple coding agents in parallel