How we build AI

The rConfig AI approach

We build AI that stays inside your control, grounded in real network knowledge, and honest about what it does today. You bring the model. Your configurations never leave the building.

01 · The spine

The principles we build AI on

Five commitments, settled before the features. They are how we keep the build honest and your network safe.

  • P1

    One place that knows the full picture

    Every AI capability in rConfig is accountable in one place. No AI happens off to the side, nothing bolted on that the rest of the product cannot see. One home that knows what the AI can do, what it was asked, and what it answered.

  • P2

    Read-only, today

    In-app AI reads your configurations and your rConfig data. It does not write to a device or change your network. Where that goes next, and how we stage it, is set out in the network AI maturity model.

  • P3

    You bring the model, we never host it

    rConfig does not run AI models for you. You point it at a commercial provider with your own keys, or a local LLM inside your own environment. The model is yours, and so is everything it sees.

  • P4

    Governance from the start

    Access control over who can use AI, provider control over which model and endpoint are permitted, and an audit trail of prompts and responses. Light today, built to grow as the capability does.

  • P5

    One approach, two stories

    AI inside rConfig for the operator, and rConfig exposed to your own external AI tools. The same stance runs through both, and it is the same accountable place behind each.

02 · Data sovereignty

Your configs never leave the building

This is the part most AI tools get backwards. They take your configurations to their cloud. rConfig does the opposite. The intelligence comes to your network, not the other way around.

You bring your own model. For security-conscious and air-gapped teams, a self-hosted local LLM means the AI never makes an outbound call, and no configuration data ever reaches a third party. The feature works the same whether you run a frontier model on your own keys or a small local model on a server in the rack.

See how bring your own model works in AI config analysis

Your keys, or your own model

A commercial provider with your own API keys, or a local LLM running entirely inside your environment. Treated as equals.

Nothing hosted by rConfig

We do not run models for you and we do not sit in the path. No SaaS AI tax, no data harvested, no shadow copies.

Air-gapped friendly

Point it at a local endpoint and the AI works with zero outbound. The configs stay on the box, and so do the prompts.

03 · Transparency

Honest about where we are

We tell you exactly what the AI does today and what it does not. Read-only is a deliberate choice, not a gap we are quietly hoping you will not notice. An assistant that reads and explains, grounded in real network knowledge, earns trust before it is ever allowed to act.

For readers who want the staged framework, where each step sits and what unlocks the next, we set it all out plainly.

Read the network AI maturity model
04 · Grounded

Grounded, not guessing

A general model on its own guesses at syntax. rConfig grounds the AI in real networking domain knowledge, in the actual config in front of it, and in your own conventions and standards. That is the difference between an answer that sounds right and one that is right for your network.

The result behaves like a network engineer who already knows your estate, rather than a chatbot meeting it for the first time. The grounding is the product. The model is just the engine.

05 · One approach, two stories

Two ways the same stance shows up

The features live on their own pages. Here is only where each one sits in the approach.

Story one · Inbound

AI inside rConfig, for the operator

Ask your config estate questions in plain language, right where you already work. Understand a config, explain a difference, narrate a compliance finding, without leaving the screen you are on.

AI config analysis

Story two · Outbound

rConfig, ready for your AI tools

Point your own copilots and agents at rConfig. A documented, secure surface means external AI understands and drives rConfig without bespoke integration work.

MCP server
06 · Where this is going

Start where it is safe, climb in the open

We are not asking you to believe in autonomous networks to get value from network AI today. Start somewhere safe, prove it, and let each step earn the next. We will tell you exactly where we are at every stage, because the engineers we build for would accept nothing less.

You can follow the path on our rConfig AI roadmap.

AI you can actually trust with your network

Private by design, grounded in real network knowledge, honest about what it does today. Talk to us about putting it to work, or see exactly where we are on the ladder.

We use cookies

We use cookies to ensure you get the best experience on our website. For more information on how we use cookies, please see our cookie policy.

By clicking "Accept", you agree to our use of cookies.

Learn more