project44 Centralizes Control of MCP Server Access & Governance

project44 streamlines their path to an AI-native future by ensuring every agent-to-tool action is secured and approved.

Project44 is a decision intelligence platform that uses AI agents to help its clients optimize their supply chains. With AI at the core of company operations, project44 employees were early adopters of MCP servers, including for internal systems and tools. As the count of MCP servers grew, so did the complexity of integrations and governance. It quickly became clear that project44 needed to centralize control of connectivity, identity, security, and governance. The team tasked with making this happen wanted to ensure they built on an MCP platform they could shape, own and run on their infrastructure. The choice of what to build on was critical to their greater AI enablement and productivity KPIs across the greater business. That’s why project44 turned to Stacklok.

Building governance into every MCP connection

A first step for project44 was ensuring it was simple for anyone building or adding an MCP server to apply consistent authentication and governance. And all MCP servers across the organization needed to reside in one central repository with access controls and an audit trail. 

Stacklok gave project44 a shared authentication and governance layer with the right controls. Now project44 can manage access to MCP servers, capture activity for audit trails, filter the tools exposed to different clients, and modify tool descriptions without adding those capabilities to every individual server.

One platform for internal and remote MCP servers

Stacklok’s ability to apply consistent authentication and governance spans all of the internal and remote MCP servers used by project44 teams. 

Internally developed MCP servers are deployed directly through Stacklok using Kubernetes custom resources. The team also uses Stacklok to proxy remotely hosted MCP servers from third-party services. Every MCP server is containerized and there is a consistent authentication mechanism across both deployment models. 

As a result, project44 has a unified operational model for MCP regardless of whether a server was built internally or provided by an external platform. And that means the company can use third-party MCP services without giving up control over authentication or the tools employees and agents can access.

Kubernetes-native infrastructure for an AI-native company

The kicker for project44 was Stacklok’s deep expertise with Kubernetes, and Kubernetes-native design. Project44 already operates a Kubernetes-based infrastructure, and now can manage MCP servers the exact same way they manage all other workloads; the same model configuration, security, and controls are all seamlessly integrated. MCP is simply part of the foundation that the platform engineering team already uses, so Stacklok was fast to deploy and easy to manage due to the ability to extend those same controls. 

Creating the foundation for AI across project44

Stacklok is helping project44 turn MCP from a collection of individual integrations into shared infrastructure for AI development and enablement across the business.

The company can now:

• Deploy internally developed MCP servers through a consistent Kubernetes-native platform
• Support coding agents and broader AI productivity initiatives with governed access to enterprise systems
• Have one control point for the way AI interacts with organizational data
• Proxy remote MCP servers through the same operating model
• Apply centralized authentication without building it into every server
• Audit which users and agents call individual tools
• Control the tools exposed to different clients

For project44, this foundation supports a larger transformation. The company is not treating AI as a standalone feature or isolated productivity tool. It is working toward agent-driven experiences across its products while enabling its engineering organization to build and operate with AI more broadly.

Stacklok gives its platform team the governed, Kubernetes-native MCP layer required to move that vision forward.