At AGNTCon Europe, ensuring AI agents don’t kill us all

At AGNTCon + MCPCon Europe 2026 in Amsterdam, attention centered on agentic artificial intelligence (AI), but there were no fears of AI causing widespread harm. Frontier model vendors, without booths at the event, seemed to emphasize global magnanimity while startups displayed their products in the exhibition hall. The focus was on agentic AI entering the market, with innovative companies showcasing their offerings.

Despite the fast-paced innovation outpacing adoption, there were genuine accounts of customers paying for these services. However, the market remained in its infancy, with numerous unproven concepts and products. I interviewed several exhibitors and identified seven companies that exemplified the conflict between rapid innovation and actual customer acceptance: Traefik Labs SAS capitalized on its cloud-native proxy and ingress controller technology to create an API, AI, and Model Context Protocol (MCP) gateway gateway.

This combined product offers a unified control plane to monitor and manage AI agents' interactions with their environment, providing necessary evidence of behavior for regulatory and forensic purposes. Bluerock Security Inc. built upon its military-grade microkernel hypervisor, resulting in a controlled AI workspace designed to protect developers and business users from unauthorized access to AI models and ensure adherence to enterprise security regulations.

Both companies provide tools to control agentic code scanning and manage token usage. Orkes Inc., born from Netflix Inc., brings controlled orchestration to agentic AI through its experience in scalable cloud-native orchestration. Orkes enables organizations to build deterministic orchestrations, ensuring less likely failures and offering explainability to trace agentic interactions.

Grape Up Sp. Z o.o. introduced Aiboostr, an AI orchestration and governance platform catering to EU Artificial Intelligence Act requirements. The act mandates organizations maintain inventories of AI capabilities, track agents' owners, purposes, and users, tasks efficiently handled by Aiboostr's built-in catalog. Two vendors showcased LLM plugin capabilities at AGNTCon: Manufact Inc. offers a cloud-based MCP platform for developers to build and run components, including LLM plugins, with analytics, testing, and compliance tools for production deployment.

Alpic SAS employs LLM plugins to provide users with functional mini-applications through their AI chat interfaces, enabling dynamic interactions and seamless booking processes. The presence of these companies at AGNTCon + MCPCon Europe 2026 highlights the dual focus on innovation and control within the rapidly evolving agentic AI sector.