January 2026 Summaries

MCP Gateway addresses the complexities of managing Multiple Control Protocol (MCP) servers by providing a centralized platform for authentication, access control, policy enforcement, and observability, facilitating easier and more secure management of AI agents interacting with company systems. As MCP adoption increased, issues such as authentication chaos, lack of access control, and difficulty in understanding tool usage emerged, leading to the development of the MCP Gateway, which acts as a control plane for managing these operational concerns. By authenticating once through the gateway, agents can access various MCP servers without managing separate credentials, while access control is enforced at multiple levels, allowing for flexible permissions without redeployment. The gateway also offers a registry for tracking server deployments and tool usage, and applies policies to ensure compliance and security across all interactions. This infrastructure, built on robust production systems, is open source, enabling organizations to adopt and customize it without vendor lock-in, ultimately making MCP infrastructure scalable and operable across entire organizations.

In the complex landscape of Large Language Model (LLM) pricing, various providers employ different metrics such as tokens, characters, and compute units, leading to confusion and the need for constantly updated spreadsheets to track costs. This inconsistency is compounded by issues like varying model names, disparate unit charges, hidden surcharges, and frequent price changes that make maintaining accurate cost data challenging. While there have been efforts like Theo's GitHub Gist, LiteLLM's Pricing JSON, and OpenRouter's API to address these issues, they each have limitations such as manual updates, lack of APIs, or authentication barriers. Portkey aims to solve this by open-sourcing a comprehensive and dynamic database that tracks pricing and usage data for over 2,000 models across 40+ providers, offering a free API without authentication and a community-maintained GitHub repository, ensuring accuracy and accessibility for enterprises relying on LLMs for significant daily operations.

Fontys ICT, a university of applied sciences in the Netherlands, developed a gateway-based AI platform to address challenges associated with the rapid adoption of generative AI in higher education. The six-month pilot involved 300 users and aimed to ensure fair access, cost control, and compliance with European law while integrating multiple AI providers. Fontys ICT's approach separated user experience, governance, and model providers, allowing the institution to maintain control without relying on a single vendor. The platform employed a layered architecture, including a frontend interface, a gateway for policy enforcement, and a diverse set of model providers. This design ensured centralized governance, allowing users to consciously select models while reinforcing educational outcomes. Key learnings from the pilot included the importance of continuous AI governance, the effectiveness of default EU-first routing, and the role of budget controls in promoting responsible usage. The platform demonstrated that universities could balance innovation and governance, offering advanced AI capabilities without sacrificing oversight.

LLM hallucinations, which occur when language models generate outputs that are coherent but factually incorrect or ungrounded, represent a significant reliability issue in production systems, as they can lead to incidents when models operate beyond their tested assumptions. These hallucinations become more pronounced with scale and complexity, as real systems introduce variability and require models to handle unpredictable, complex user inputs and longer prompts. Notably, model upgrades alone are insufficient to eliminate hallucinations, as orchestration decisions and system-level controls play critical roles in managing these issues. The introduction of an AI gateway can mitigate hallucination risks by enforcing consistency, making model selection explicit, and constraining tool usage and side effects. Additionally, integrating guardrails alongside an AI gateway is essential to validate outputs against explicit rules, providing a feedback loop that connects real production interactions with upstream changes, ensuring hallucinations are treated as debuggable system behaviors rather than isolated incidents. These measures help prevent hallucinations and their potentially harmful impacts, ensuring more reliable and coherent AI system outputs in production environments.