Project Background

As the use of artificial intelligence grew at Lastlink, different teams began developing agents, automations, and integrations with internal tools. While these initiatives showed significant potential, each solution could end up implementing its own credentials, access rules, and methods of communicating with the company’s services. I realized that scaling AI adoption securely would require shared infrastructure. Based on this need, I designed and developed Lastlink’s AI Gateway: a central platform connecting AI applications to the organization’s internal tools.

Solution Development I built the project from the ground up, contributing to its strategy, architecture, implementation, and production deployment. The solution was based on the Model Context Protocol (MCP), which provides a standardized interface through which AI agents and applications can discover and use different tools. The gateway acts as an intermediary layer between AI agents and internal services. Applications do not need to understand the implementation details of every integration. Instead, they communicate through a shared interface while the platform manages access to and orchestration of the available tools.

The infrastructure was deployed on Kubernetes and designed for production requirements such as observability, failure handling, security, and continuous evolution.

Governance and Security

One of the project’s main goals was to ensure that AI adoption took place in a governed environment. Authentication, access policies, and usage boundaries were therefore considered from the beginning of the architecture.

This approach reduces credential exposure, prevents fragmented integrations, and provides greater control over which tools each application or agent can access. Standardization also makes it easier to introduce new tools without requiring every team to build a separate integration.

Technologies and Responsibilities

For this project, I was responsible for AI strategy, system architecture, platform implementation, MCP-based orchestration, Kubernetes deployment, and supporting adoption across teams. Due to confidentiality requirements, specific details about the internal architecture, connected tools, and security policies are not included in this case study.