Building Sierra's MCP Gateway: An engineering iceberg (sierra.ai)

Sierra has announced the development of its new MCP Gateway, a central hub designed to safely connect AI agents to numerous tools and systems across the company. This initiative arose from the need to ensure that their internal agent, Pinecone, could access various types of contextual data seamlessly from platforms like Slack, GitHub, Salesforce, and more. The gateway is powered by the Model Context Protocol (MCP), which offers a standardized method for enabling these integrations, despite being a relatively old technology by industry standards. By consolidating access through a single gateway, Sierra avoided the pitfalls of fragmented systems and varying permissions across departments, leading to streamlined operations and improved productivity. This project highlights significant lessons learned by Sierra's AI Acceleration team, underscoring the importance of clear ownership in engineering roles, comprehensive documentation, and meticulous data protection strategies. They implemented a robust auditing mechanism to prevent unauthorized access to sensitive customer information, an effort that reassured legal and compliance teams. The team also recognized that to maximize user adoption, the gateway must effectively support entire workflows, adapting to user requirements rather than just offering partial solutions. Overall, the MCP Gateway exemplifies how a unified AI framework can enhance efficiency and agility in an AI-first engineering organization, paving the way for smarter, context-aware automation.