How can we enhance this page for you?
We are constantly evolving. Your feedback helps us build more intuitive and informative product pages.
Agentic AI Platform on Kubernetes
A sovereign, open-source platform for generative and agentic AI on STACKIT. A central gateway ensures access control, cost budgets, LLM failover, guardrails, and tracing for every agent, prompt & MCP call.
Highlights
Sovereign and open source
Built entirely on Kubernetes and STACKIT, using established open-source components instead of proprietary lock-in. Your data and AI workloads stay under your control at all times – sensitive prompts never have to travel through external hyperscaler APIs.
Governed from day one
A central AI gateway unifies access to all models and MCP tools. SSO, RBAC, token budgets, guardrails, and tracing are enforced centrally – and inherited automatically by every new agent, instead of being rebuilt for each project.
In production in days, not months
A strict "Golden Path" template and GitOps delivery bring new AI agents into production in days, without deep Kubernetes expertise. Your developer teams work in self-service, within clearly defined guardrails.
Predictable costs, no per-token pricing
Open-weight LLMs can optionally be self-hosted directly on STACKIT – only GPU capacity is required, with no dynamic per-token pricing. Token budgets and per-user/per-agent cost attribution provide full transparency.
Overview
Vendor
Delivery method
Professional Service
Categories
DevOps, Infrastructure Software, Machine Learning + 1 more categories
Product description
Building an AI prototype is straightforward. Running generative and agentic AI enterprise-wide - securely, sovereignly, and under cost control - is a different challenge entirely. That's why you need a centrally governed platform: an AI gateway as the single point of access to models and MCP tools, with self-service for your teams within clear guardrails for security, cost, and governance. What You Get Rather than starting from scratch, PRODYNA's experts work with your team over 10–12 weeks to adapt our proven, open-source platform blueprint to your requirements and deploy it directly on STACKIT. Contents: - Hardened Kubernetes cluster (optionally with GPU nodes) - Central AI & MCP gateway with SSO, RBAC, token budgets, guardrails, and cost allocation per user/agent - Chat UI (LibreChat) and observability with Grafana stack - "Golden Path" template and GitOps delivery — every agent automatically inherits governance, budgets, and tracing - Optionally self-hosted open-weight LLMs without per-token pricing Prerequisites: - STACKIT organization incl. Foundation (Landing Zones) - Availability of your experts (network, IAM, security) - A small group of motivated early adopters