Kong unveils AI Connectivity roadmap for unified control

Kong has set out a product roadmap for what it calls "AI Connectivity", an architectural approach to connecting and governing traffic across APIs, AI models and agent-based systems.

It described the shift as moving from connecting services to connecting "intelligence" across a wider mix of components, including models, tools, data sources, real-time streams and autonomous agents. As organisations move beyond pilots and put advanced AI systems into production, Kong said, they face growing operational complexity.

At the centre of the plan is Kong Konnect, its cloud platform for API management and related tools. The AI Connectivity roadmap aligns with Konnect and recent product releases across gateways, testing, governance and billing.

Unified governance

Kong described AI Connectivity as a way to apply a single governance and runtime layer across multiple kinds of traffic. It included APIs, events, large language model calls, MCP connections and agent-to-agent communication. The goal is consistent visibility, control and policy enforcement across what it calls the "agentic stack".

Many organisations now run AI workloads alongside existing API estates, adding new control points and routes for sensitive data. AI agents can also interact with tools and services on a user's behalf, expanding the set of systems that require monitoring and access control.

Kong positioned its approach around three blockers to scaling AI in production: speed, cost and risk. It said AI Connectivity unifies the "full AI data path" across systems so that intelligence can flow "securely and reliably".

"As organisations shift from connecting services to connecting intelligence, enterprises need a foundation for how AI systems are built and operated," said Augusto Marietti, CEO and Co-Founder of Kong. "With AI Connectivity, Kong is defining a new vision for how APIs, AI models and agents can work together in a unified, enterprise-ready architecture that enables companies to scale AI safely, cost effectively and with confidence."

Product roadmap

Kong tied the roadmap to several products in its portfolio, including Kong AI Gateway, the Insomnia API client, Konnect Metering & Billing, and KAi, an assistant within Konnect.

It highlighted governance features for MCP, a protocol used to connect AI models and agents to tools and services. It also pointed to MCP testing support in Insomnia and new controls in Konnect Metering & Billing that it described as cost governance and monetisation features.

Kong said the combined toolset covers routing and governance for model traffic, discovery of tools used by agents, real-time streaming, and observability across an organisation's deployments.

MCP registry

A new element of the roadmap is an MCP Registry, which Kong said will sit inside Kong Konnect's API Service Catalog. It described the registry as a way for enterprises to register MCP servers and tools, support dynamic discovery, and apply governance for AI agents that use those services.

Tool discovery has become more important as enterprises experiment with agent-based designs. In these systems, an agent can choose from a menu of tools and connect to external services as it carries out tasks. Organisations often want to track which tools are available, who can use them, and what data flows through them.

Operational focus

Kong argued that moving from experimentation to production depends on operational controls. It said a unified approach can reduce integration work when connecting APIs, models, agents and tools, while centralised governance and visibility can reduce compliance and operational risk by limiting "shadow AI".

Cost is also a growing concern, particularly where teams struggle to understand and control usage across multiple models and applications. Kong said its approach provides visibility into model usage, token consumption and API traffic, with the aim of identifying inefficiencies and managing spend.

It also said routing AI traffic through a single connectivity layer can improve reliability for real-time, agent-driven workloads, with lower latency and more consistent control.

Kong presented the roadmap during a live stream led by Marietti, Marco Palladino, Co-Founder and Chief Technology Officer, and Reza Shafii, SVP of Product Management. The session also included guests from Mistral AI and McKinsey's technology practice.

"At Mistral AI, our mission is to put AI in everyone's hands. We're focused on building the full stack-from the models to applications to compute. Putting AI to work at scale is critical to our day to day, and that's one of the reasons why partners like Kong are important," said Lélio Renard Lavaud, VP of Engineering at Mistral AI. "Through their work in AI connectivity, they help reduce friction and siloes-serving the industry more broadly."

"Individual pilots only get you so far in the enterprise. Accelerating innovation with AI requires deployment at scale, supported by platforms that leverage enterprise-grade services. Companies need a way to connect and govern APIs, models and tools to use AI securely and efficiently at scale," said James Kaplan, Chief Technology Officer, McKinsey Technology Practise.

Kong said its next steps include rolling out the MCP Registry within Konnect's service catalogue and continuing to expand governance, testing, metering and assistant features under the AI Connectivity roadmap.