Omnigent is an open-source AI agent framework and meta-harness with more than 10,900 stars on GitHub that orchestrates Claude Code, Codex, Cursor, Pi, and custom agents behind one governed, collaborative layer.
Every serious team now runs several coding agents, and each one has its own harness, its own configuration style, and its own idea of safety. Omnigent sits above them all as a meta-harness: you define your work once, then swap the underlying agent — Claude Code today, Codex tomorrow, a custom in-house agent next quarter — without rewriting prompts, policies, or workflows. Governance is a first-class feature rather than an afterthought: policies and sandboxing constrain what agents may touch, so autonomy can be increased without losing control. Collaboration is built in too, with real-time multi-user sessions that let a team watch, steer, and contribute to agent work from any device, phone included. The framework is Python-based, Apache-2.0 licensed, and designed for organizations standardizing on agents the way they once standardized on CI: one platform, many engines, consistent guardrails. For individual developers it is equally useful as a way to stop re-learning a new harness every time a better agent ships.
Omnigent is completely free and open source under the Apache-2.0 license. You run it yourself and bring your own agent subscriptions and API keys; there is no paid tier for the framework.
Engineering teams and platform builders standardizing on multiple coding agents, who need portability between harnesses, enforceable policies, and shared visibility into what their agents are doing.