Kagenti provides a cloud-native, framework-neutral middleware for deploying and orchestrating AI agents in production, addressing challenges like deployment complexity, security gaps, and protocol fragmentation. It enables reliable, scalable, and secure operation of diverse agent frameworks using standardized A2A and MCP protocols.

How It Works

The platform is structured around four pillars: Lifecycle Orchestration, Networking, Security, and Observability. It enhances existing agent frameworks (e.g., LangGraph, CrewAI) with production-ready infrastructure. Key components include the Kagenti UI for management, a Workload Runtime, and platform services. These manage agent/tool lifecycles, A2A/MCP networking via an MCP Gateway and service mesh (Istio/Ambient), zero-trust security with SPIRE for workload identity, and observability features. This approach decouples agent logic from deployment concerns, promoting consistency.

Quick Start & Requirements

Requires Python ≥3.9 (uv recommended), Docker/Podman (16GB RAM, 4 cores recommended), Kind, kubectl, Helm, and Ollama. Installation involves cloning the repo, checking out a release (e.g., v0.6.0), configuring secrets, and deploying to Kind using scripts/kind/setup-kagenti.sh --with-ui --with-spire --with-agent-sandbox --with-builds. Access the UI at http://kagenti-ui.localtest.me:8080 after running show-services.sh. Detailed OpenShift instructions are in the Installation Guide.

Highlighted Details

• Zero-Trust Security Architecture with SPIFFE/SPIRE for trusted workload identity.
• Framework-neutral support for diverse AI agent frameworks via standardized A2A protocol.
• Unified MCP Gateway for Model Context Protocol tools, enforcing policy and routing.
• Kagenti UI dashboard for deploying, testing, and monitoring agents/tools as Kubernetes Deployments.
• Leverages Kubernetes primitives, including AgentCard CRD for lifecycle management.

Maintenance & Community

Contributions are welcomed via CONTRIBUTING.md. Contact maintainers at kagenti-maintainers@googlegroups.com or join the Slack community.

Licensing & Compatibility

Licensed under Apache 2.0, permitting commercial use and integration into closed-source projects.

Limitations & Caveats

The provided documentation does not explicitly detail limitations, alpha/beta status, or known bugs. Core architecture relies on Kubernetes/OpenShift environments.