Coveo and Kong Launch Tools to Govern AI Agents
Two companies have released platforms to help enterprises connect data to AI agents in a governed manner. Coveo announced a hosted server to standardize and govern the use of AI agents across an organization. Separately, Kong launched Context Mesh, a solution designed to make existing company APIs ready for consumption by AI-driven workflows.
- Both Coveo's Hosted MCP Server and Kong's Context Mesh are built upon the Model Context Protocol (MCP), an emerging standard for how AI agents discover and interact with external tools and data sources. This protocol-based approach aims to prevent vendor lock-in and create a more interoperable AI ecosystem. - Kong's Context Mesh automatically discovers existing enterprise APIs and transforms them into toolkits that AI agents can consume, reducing what could be weeks of manual API inventory and mapping effort. This allows companies to reuse their existing API investments for new AI applications instead of building from scratch. - Coveo's solution focuses on providing secure access to enterprise content, connecting its unified index of data sources to large language models like ChatGPT Enterprise and Anthropic's Claude. An application is available in the ChatGPT Enterprise directory to act as a secure bridge to content indexed by Coveo. - A key challenge these tools address is that enterprise data is often scattered across various APIs, applications, and event streams, each with different formats and access rules. Manually integrating agents with each source is slow and risky, a problem Gartner predicts could put 40% of agentic AI projects at risk of cancellation by 2027 if not solved. - Governance is a central feature, as AI agents often require elevated privileges to access databases and internal systems. Platforms like these provide observability, policy enforcement, and audit trails to monitor agent behavior, detect threats, and prevent actions like data exfiltration. - Kong's platform includes an MCP Registry where all generated toolkits are registered for enterprise-wide discovery and reuse, giving developers a central catalog for both traditional APIs and new AI agent tools. The toolkits are deployed to the Kong AI Gateway where security policies like rate limiting and authentication can be enforced at runtime. - The Coveo MCP Server provides pre-built tools for agents to perform tasks like searching indexed content, retrieving specific passages, and generating natural-language answers with citations using its Relevance Generative Answering API. - These platforms represent a shift from Retrieval-Augmented Generation (RAG), where an LLM is simply fed external data, to a more dynamic "agentic" architecture where autonomous agents can reason, plan, and invoke external tools to solve problems.
