What is graftcode?

Graftcode is a development layer that lets you call methods written in one language, running on a remote service, as if they were a local dependency. The pitch is that you write ordinary public methods, run them on a host called Graftcode Gateway, and then install an auto-generated, strongly typed client package, called a Graft, in any other project through its normal package manager, whether that is npm, dotnet, or pip. From there, calling the remote method looks like calling local code, with no hand-written API layer, no OpenAPI spec, and no client code to maintain. The project supports methods written in .NET, Java, Python, Node.js, Go, PHP, C++, and Ruby, and can run on Windows, Linux, macOS, and Android, with cloud options for Azure, AWS, GCP, and Docker. The README claims this removes the need for REST, gRPC, Thrift, or message-bus coupling between services, can cut codebase size significantly, and adds built-in MCP support so any backend becomes usable by AI agents out of the box. It also says teams can switch between a single monolith and separate microservices, or change the underlying communication channel between WebSocket, HTTP/2, RabbitMQ, Kafka, Service Bus, SQS, or Pub/Sub, without changing application code. Getting started involves packaging your library, downloading the Graftcode Gateway with a one-line install script for Windows, Linux, or macOS, and running the Gateway against your compiled module. On the consuming side, a project installs a Graft through its package manager pointed at a project-specific registry URL, then imports and calls the generated client classes directly. The README includes an interactive quick-start tutorial series covering topics like connecting a frontend to a backend, exposing your first service, and switching between monolith and microservice setups. The organization also maintains separate repositories for the core Gateway engine, live demos, full documentation, Docker images, and plugins for various messaging systems.