Skip to content
HelvetiCode
Menu
ENDE
Explore a design partnership

For software teams with regulated clients

Every AI coding request, checked before it reaches a model.

A gateway between the AI tools your developers already use and the models behind them. Designed to apply your clients' rules to every request, route it to a model you approved, and keep a record your auditors can use.

Explore a design partnershipHow it works
Illustration: HelvetiCode, Request 04821
Illustration
From
IDE · payments-core
Prompt
Use ACT/365 day count for CHF accounts
Context
interest.ts, config.tsAPI_KEY=[redacted]
  1. Secret scan1 key found, redacted
  2. Repository rulepayments-*: self-hosted only
  3. Modelrouted to self-hosted model
  4. Audit trailrecorded
DecisionAllowed after redaction

The problem

Your clients will ask where their code went.

Banks, insurers and healthcare providers set rules for how their suppliers handle code and data. With most AI coding tools, nothing sits between the developer and the model to apply those rules, or to show afterwards that they were followed.

  • Which model saw our code?

  • Did credentials or client data leave the environment?

  • Under which rule, and who approved it?

How it works

One checkpoint for every request

The gateway is designed as the single path between your developers' tools and the models. Developers keep their editor; the rules live in one place.

Your environment or Swiss-hosted

Developer tools

IDEs and coding agents

HelvetiCode Gateway

  1. Check
  2. Route
  3. Record

Approved models

Self-hosted open models; external only where allowed

Audit trail

Append-only, access-restricted

Architecture diagram: inside your own environment or a Swiss-hosted one, developer tools such as IDEs and coding agents send every AI request through the gateway. The gateway is designed to check the request, route it to an approved model (self-hosted open models, or external models only where your rules allow), and write the request and its decision to an append-only audit trail.

  1. Check

    Before a request reaches any model, it is checked for secrets, repository scope, allowed models and sensitive content. The outcome: allow, redact, reroute or block.

  2. Route

    Approved requests go to a model your rules allow: self-hosted open models by default, external models only where permitted.

  3. Record

    Each request, decision and model choice is written to an append-only audit trail, ready for your client's audit or your own.

Read the architecture

Who it is for

Built for teams whose clients set the rules

  • Software suppliers to banks and insurers

    Show a client which model saw their code, and under which rule.

  • Healthcare, medtech and pharma software

    Keep regulated code and data on models you have approved.

  • In-house teams in regulated organisations

    Give developers AI assistance that your risk function can sign off.

Runs where you decide

Designed to run on-premise or Swiss-hosted, with self-hosted open models, so code stays inside the environment you approved.

Design partnership

What's on your mind?

We are shaping the product with a small number of software teams. If your clients set rules for how you use AI, we would like to hear about them.

Let's talk

Email contact@helveticode.dev

Useful to include

  • Who your clients are and which rules they set
  • Which AI coding tools your team uses today
  • What you would need to show in an audit