Turn any service into MCP tools - Released EE v2.12

MCP is an emerging standard designed to unify how AI clients (Such as chat interfaces, code editors, and others) interact with services. If you already have services (such as REST endpoints, gRPC, Lambda functions, etc.), KrakenD can handle bidirectional communication without requiring any changes to your backends.

The new release of KrakenD EEv2.12 delivers a bold step forward towards rapid AI adoption in an impressive time-to-market and without requiring new tools, as it converts a wide range of protocols and encodings to MCP. To illustrate what we are talking about, let’s start with a (maybe extreme) scenario:

Your company has a 15-year-old XML-based service using the SOAP protocol that no one dares to touch. There’s no way you can have an AI Agent talking to it, right? Well, KrakenD can enable that for you.

By adding the MCP Server component to your KrakenD configuration, you can expose the SOAP service to AI Agents. Imagine an interaction like this on Claude that used data from the SOAP service:

Image: Screenshot of an interaction of Claude with KrakenD MCP Server feeding from a SOAP server

This release revolutionizes how you can build API gateways that speak the language of modern AI ecosystems, whether bridging legacy systems or orchestrating complex service workflows.

While other API Gateways originate from a proxy-only base and struggle to embrace MCP, this came naturally to KrakenD, as it has always had the power to understand business logic.

Introducing the MCP Gateway and the MCP Server modes

The MCP functionality on KrakenD allows you to quickly implement two scenarios:

1. Use KrakenD as the MCP Gateway, proxying AI clients that speak MCP to your existing MCP servers
2. Start an MCP Server to autogenerate MCP tools within KrakenD itself

This functionality adds all the power of KrakenD, allowing you to do crazy stuff. Examples:

Do you want Claude to consume a legacy SOAP XML service? Check. Want to convert your traditional REST API to MCP tools? Check. Want to plug a conversational chat into an AWS Lambda? Check. Agent to Queue? Check. Trigger a workflow from an IDE? Check. Aggregate multiple services and offer them as a unified tool? Check. Consume gRPC from the AI? Check. Govern AI usage and throttling? Check. Monitor in Grafana all the things? Check. And the party goes on and on..

For more information on the new MCP functionality check the KrakenD MCP documentation below:

Resilience and observability improvements

In addition to MCP, we have released a new HTTP Circuit Breaker that lets you define which HTTP status codes should trigger circuit-breaking, fine-tuning fault tolerance behaviors for your APIs.

Also, the OpenTelemetry tracing supports skipping headers that might contain sensitive information, boosting security and privacy in observability data even that is for internal usage.

Security and configuration updates

The Security Policies engine has been upgraded for more robust access controls. We replaced the configuration parser to handle case sensitivity more predictably, especially in OpenAPI schema parsing, which is now case sensitive to facilitate better validation and integration.

Developer and platform enhancements

We now offer ARM64 Docker images for Enterprise plugin builders, improving support for diverse infrastructure without needing to cross-compile. Moving forward, all releases will continue to include the ARM64 option in addition to AMD64. Images available on Docher Hub.

Happy AI building! 🐙