What goes to Graftcode Cloud

One of the most important questions teams ask when evaluating Graftcode is simple and reasonable:

Note

What data leaves my environment?

Graftcode is designed so that this question has a clear, precise, and verifiable answer.

The short version is:

  • interface metadata goes to the cloud
  • execution and runtime data NEVER DO

This section explains exactly what that means in practice.


The role of the cloud in Graftcode

The Graftcode cloud exists to support development-time workflows.

Its responsibilities are limited to:

  • enabling on-demand Graft generation
  • supporting package manager requests
  • providing project-level visibility and dashboards (when enabled)

The cloud is not involved in production execution.

Understanding this separation is key to understanding the security model.


What is sent to the cloud

When a Graftcode Gateway starts in project-bound mode, it analyzes the modules it hosts and builds a Unified Graft Model (UGM).

Only the following information is sent to the Graftcode cloud:

  • public class names
  • public method names
  • method signatures
  • argument and return types
  • structural information required to generate typed clients

This information describes what can be called, not how it is implemented.


What is explicitly not sent

The Graftcode Gateway does not send:

  • business logic
  • source code
  • compiled binaries
  • runtime data
  • method arguments
  • return values
  • exceptions thrown during execution
  • configuration secrets
  • environment variables

At no point does production traffic or execution data leave your environment.


No production execution in the cloud

Once a Graft is generated:

  • calls go directly from caller to receiver
  • execution happens through Hypertube
  • communication stays within your network

The Graftcode cloud:

  • does not forward calls
  • does not observe payloads
  • does not participate in execution
  • does not sit in the data path

From a runtime perspective, production systems do not depend on the cloud.


Package manager requests and client generation

When a package manager requests a Graft:

  • the request reaches the Graftcode Engine
  • the Unified Graft Model is used to generate a client
  • the generated package is returned to the requester
  • the result is cached for efficiency

The cloud sees:

  • which interface was requested
  • which language and version were requested

It does not see:

  • how the client will be used
  • where it will be deployed
  • what data it will process

Metrics and dashboards (optional)

When a Graftcode Gateway is linked to a project, it may optionally report:

  • high-level usage metrics
  • interface-level activity counts
  • version adoption information

These metrics are:

  • aggregated
  • non-payload
  • configurable

They exist to help teams understand:

  • which interfaces are used
  • which versions are still active
  • where deprecations are safe

All detailed logging and observability remains local and under your control.


Security boundaries and network control

From a network perspective:

  • all production execution stays inside your network
  • all communication uses standard transports
  • existing firewalls, proxies, and load balancers apply unchanged

There are no hidden connections or backchannels.

If your infrastructure can enforce network boundaries today, it can enforce them with Graftcode.


Verifiability and transparency

The Graftcode Gateway is intended to be source-available, allowing teams to:

  • inspect exactly what it does
  • verify what data it sends
  • perform independent security analysis and penetration testing

This design ensures that security does not rely on trust alone, but on inspectable behavior.


Why this model exists

Graftcode separates concerns deliberately:

  • interfaces are shared
  • execution remains local
  • control stays with the user

This allows teams to benefit from:

  • strongly typed integration
  • on-demand client generation
  • runtime-level execution

Without surrendering control over their systems or data.


In short

The Graftcode cloud:

  • receives only interface metadata
  • enables client generation and discovery
  • is not part of production execution

Your code:

  • runs in your environment
  • communicates directly
  • remains under your control

This clear boundary is foundational to the Graftcode security model.


See also: How Grafts are generated