The Core
Lightweight, event-driven Flask & Alpine.js runtime running Python 3.13 loops. Designed for continuous self-improvement (RSI).
Agent Zero is the open-source execution framework for sovereign multi-agent workflows. Deployed locally or on GKE, Tailscale-meshed, and run with No Human In Loop (NHIL). A standard for developers who refuse legacy middleware.
Run complex terminal commands, file edits, and multi-agent loops directly on your machine or inside isolated Docker containers. Built for speed, safety, and absolute control.
We build the open-source autonomous standard. Clone the Agent Zero execution framework to orchestrate local workflows. Build and run custom plugins. Or join the developer community to shape the future of sovereign software.
Lightweight, event-driven Flask & Alpine.js runtime running Python 3.13 loops. Designed for continuous self-improvement (RSI).
An interactive, responsive workspace shell for real-time trace debugging, file management, and direct subagent dispatch.
Extensible via custom plugins and MCP servers. Built to be modular, zero-dependency, and entirely self-contained.
Python-native executor with direct terminal, file system, and subprocess access. Zero cloud dependency, fully open-source.
Extend capabilities with pre-built or custom plugins (Tailscale mesh, Keycloak auth, GCP/AWS integrations) under usr/plugins.
Eagerly load and execute Model Context Protocol servers (chrome-devtools, gke-oss, firebase) directly in agent tasks.
We design our agents to write their own skills, optimize code paths, and self-heal in response to test failures (Recursive Self-Improvement). The entire platform is built to execute without human intervention, creating a persistent, self-governing local loop.
No Human In Loop.
Where big tech locks you into expensive API subscriptions and proprietary clouds, KGU Open Source puts the tools back in your hands. Run models locally, connect your own tools, and own your identity.
Partnered with the JC Coffey Foundation · "NO HUMAN LEFT BEHIND"
Get started in seconds. Install dependencies, run initialize scripts, and spin up your first autonomous agent locally.
Github Repository →Write customized tools, prompt templates, and agent profiles. Share them with the community or run them locally in usr/plugins.
Developer Guide →Test prompts, run sandbox tasks, and experiment with model behaviors directly in your browser.
Open Playground →