Agentex provides an open-source framework for building, deploying, and scaling AI agents across five capability levels (L1-L5), addressing the limitations of traditional synchronous AI applications. It targets developers and enterprises seeking to create advanced, autonomous, and future-proof AI workflows, offering a unified, cloud-agnostic architecture that supports seamless growth from basic chatbots to complex agentic systems.

How It Works

The core approach enables agents to move beyond simple request/response patterns, supporting complex, long-running, or autonomous workflows. Agentex utilizes a Python SDK and a flexible, cloud-agnostic, Kubernetes-native architecture. Agents are hosted and scaled independently, allowing developers to build from L1 (basic) to L5 (fully autonomous) capabilities without altering the fundamental architecture, facilitating a smooth progression in agent sophistication.

Quick Start & Requirements

• Primary Install: uv tool install agentex-sdk (globally available).
• Prerequisites: Python 3.12+, uv package manager, Docker, Node.js. Local Redis may need to be stopped.
• Setup: Involves starting the Agentex Server (make dev in agentex/) and the Frontend Server (make dev in agentex-ui/), then initializing an agent (agentex init) and running it (agentex agents run --manifest manifest.yaml). Requires setting an API key (e.g., OPENAI_API_KEY) in a .env file.

Highlighted Details

• Supports AI agent capabilities across five levels (L1-L5), from chatbots to fully autonomous systems.
• Enables seamless progression from basic to advanced agentic AI without core architectural changes.
• Features a development UI and agent server for local development and testing.
• Cloud-agnostic and Kubernetes-native architecture.

Maintenance & Community

• Maintainers: @danielmillerp, @RoxyFarhad, @smoreinis, @MichaelSun48, @declan-scale.
• Original Authors: @felix8696, @jasonyang101.
• Community support is available via GitHub issues, discussions, and pull requests.

Licensing & Compatibility

• The README describes an "Open Source Edition" but does not explicitly state the license type.
• Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The open-source edition requires manual deployment and lacks enterprise features such as GitOps setup, integrated builder tools, AgentOps lifecycle management, and centralized identity management. Local development has specific prerequisites and potential conflicts with existing services like Redis.