AI agents are already calling your APIs
Your teams connect LLMs and agents to internal systems. Each connection is an API call. Many go around the controls you built for people and applications.
The Model Context Protocol (MCP) makes these connections easier. It also multiplies them. Without a gateway, you lose track of who calls what, at what cost, and with which data.
Kong puts APIs, LLM traffic and MCP tools behind one control point. Same policies, same identity, same logs.
What we deliver on Kong
API gateway. Design, deploy or migrate Kong Gateway, self-hosted or on Kong Konnect. Routing, authentication, rate limiting and versioning, set up as code.
API platform and governance. Set up Kong Konnect as your API platform: service catalog, developer portal, ownership and standards. Teams publish APIs faster, under the same rules.
AI Gateway. Route traffic to OpenAI, Anthropic, Azure, Bedrock or self-hosted models through Kong AI Gateway. Control access, rate limits and token costs per team or application.
MCP management. Expose existing REST APIs as MCP tools without writing new servers. Secure them with OAuth 2.1, group them so each agent only sees the tools it is allowed to use, and register approved servers in a central catalog.
Security and identity. Treat agents as identities. We connect Kong to your IAM, define least-privilege access and send logs to your SIEM. This is where our IAM and PAM practice comes in.
Run and support. Upgrades, monitoring, incident support and platform ownership, by a project team or by a consultant embedded in your team.
MCP management, in plain terms
MCP (Model Context Protocol): an open standard that lets AI agents discover and call tools, such as your APIs.
MCP server: the component that exposes those tools to agents.
MCP gateway: a control point in front of MCP servers. It authenticates agents, filters which tools they see, applies rate limits and logs every call.
MCP registry: a catalog of approved MCP servers, so agents only use tools you trust.
With Kong, the same platform manages your APIs and your MCP tools. You do not need a separate stack for AI.
How we work
1. Assess. We map your APIs, gateways and AI use cases, and identify the risks: unmanaged keys, shadow AI, missing logs.
2. Design. Target architecture, security model and governance rules, validated with your architecture and security teams.
3. Build. Deployment as code, migration of existing APIs, first MCP tools exposed to a pilot agent.
4. Run. Handover to your teams, or ongoing support by FuturWork consultants.
Engagement models: a fixed-scope project, or consultants embedded in your teams.
Why FuturWork
Kong partner. We design and deploy Kong for organizations in Belgium, France, Luxembourg and the Netherlands.
Security first. Our core practice is cybersecurity: IAM, PAM and DevSecOps. Your gateway is designed as a security control, not only as plumbing.
Local. Based in Brussels, with consultants across our markets.
Frequently asked questions
What is the difference between an API gateway and an AI gateway?
An API gateway manages traffic between applications and your APIs. An AI gateway applies the same controls to traffic toward LLMs: access, rate limits, token costs and logs. Kong does both on one platform.
Do we need to rewrite our APIs to make them available to AI agents?
No. Kong can expose an existing REST API as an MCP server. You keep your API and add MCP access with the security policies you already use.
Konnect or self-hosted?
Both work. Kong Konnect is Kong's cloud control plane, with data planes in your cloud or data center. Kong Gateway Enterprise runs fully on your infrastructure. We help you choose based on your security and data residency constraints.
Can Kong replace our current API management tool?
Often, yes. We assess your current setup and plan the migration in stages, without breaking existing consumers.
Ready to put your APIs and AI agents under control?
Tell us where you are today, and a Kong expert will get back to you. Talk to a Kong expert. Read also: AI agents in the enterprise: 7 security risks CISOs must plan for.