Demos, reference architectures, and tooling for deploying AI agents on OpenShift with Kagenti.
Part of Red Hat's Agents & AI Engineering Tools organization, the Agent Ops team builds the infrastructure for agent deployment, identity, and token exchange on OpenShift AI.
agent-ops/
├── demos/
│ └── redbank-demo/ Multi-agent banking demo (MCP + A2A + RLS)
└── README.md
A multi-agent banking application that demonstrates the Kagenti platform's core capabilities: agent enrollment, JWT-based identity propagation, AuthBridge sidecar injection, and A2A agent discovery.
What it shows:
- MCP server with auth-aware tools (FastMCP + PostgreSQL + Row-Level Security)
- A2A agents — Orchestrator, Knowledge (RAG), and Banking Operations agents built with LangGraph
- Kagenti integration — AgentRuntime CRs, AuthBridge sidecar, AgentCard-based peer discovery
- RAG pipeline — LangChain + PGVector with role-scoped document collections (KFP pipeline)
- Playground UI — Chat interface with Keycloak OIDC authentication
- Role-based access control at three tiers: AuthBridge (network), application (
@admin_only), and database (RLS)
Stack: Python, LangGraph, FastMCP, PostgreSQL 16, PGVector, Keycloak, Helm, OpenShift
See the RedBank Demo README for architecture details and deployment instructions.
| Project | Description |
|---|---|
| kagenti | Kagenti platform — Helm charts, AuthBridge, SPIFFE IdP setup |
| kagenti-operator | Kubernetes operator for AgentRuntime and AgentCard CRDs |
| kagenti-extensions | Extensions and integrations for the Kagenti platform |
- Fork the repository
- Create a feature branch:
git checkout -b feat/your-feature - Make your changes and commit with conventional commit messages
- Open a pull request against
main