> ## Documentation Index
> Fetch the complete documentation index at: https://docs.viorant.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Introduction
> Welcome to Viorant, the operating layer for AI.
## Welcome to Viorant
Viorant is the operating layer for AI: a local-first, vendor-neutral control plane for
the prompts, skills, memory, models, and agents you ship to production. Author them in
one place, capture the whole system as a single declarative file, and deploy it
anywhere, reproducibly.
It is built for the team shipping AI that has to work when it matters, and cannot afford
to guess what is actually running in production.
Author and sign your AI assets on your own machine, on your own keys. Nothing leaves
your environment unless you ship it.
Portable across frameworks and clouds. One artifact runs the same everywhere, with no
lock-in to a single provider.
Prompts, skills, memory, models, and agents captured as one file you can diff and roll
back.
A real diff and a real approval before any change reaches a user.
Signed provenance on every asset: what shipped, who authored it, and that it was not
altered.
***
## Start with the Hub
The **Viorant Hub** is the local-first desktop app where you author, run, and version
your agents today — free, on your own machine, with no API key required to start.
Download, install, and open Viorant on macOS or Windows.
Sign in with Google and get to your first agent.
Test a prompt in minutes — on included credits or your own key.
Agents, the Masters, Skills, Connectors, and Models.
***
## The Vio File
The direction Viorant is building toward is the `.vio` file: your whole AI system as one
signed, declarative artifact you can version, diff, and rebuild anywhere. Terraform, but
for AI.
```yaml support-agent.vio theme={"dark"}
name: support-agent
version: 4.2.0
signed_by: alex@acme.com # verified
model: claude-sonnet-4.5
prompt: ./prompts/triage.vio
skills:
- refund.vio
- order-lookup.vio
memory: ./memory/store.vio
```
One file is the system — no guessing which version is live, no reverse-engineering a
broken deploy, no rebuild that depends on someone's memory.
Today's documentation covers the current Viorant Hub. The `.vio` format, signing, and
deployment arrive with upcoming releases.
***
## Quick Start
New to Viorant? These steps take you from install to a running agent.
Download [Viorant Hub](/getting-started/installation) for macOS or Windows (Linux is
on the way). It is free, and every account starts with credits so you can run
without bringing an API key.
Build your prompts, skills, models, and agents in the Hub, in one place.
Run your agent locally — on included credits or your own keys — inspect the full
trace, and refine.
Commit each change, review it like code, and keep a traceable history of what shipped.
***
## Core Concepts
Viorant organizes every AI system around a small set of building blocks.
The instructions that drive behavior, versioned and reviewable like code.
Reusable capabilities defined once, with a single source of truth across projects.
Model selection and routing, pinned to the version you actually shipped.
Composed systems that bring prompts, skills, and models together.
How the Hub reaches the external tools your skills and agents use.
Persistent context your agents rely on, captured as part of the system. Arrives with a
later release.
***
## The Platform
Where each part of Viorant stands today.
Build, author, and run your AI assets in one local-first environment. **Live now, free.**
Your whole AI system as one signed, declarative artifact. **In active development.**
Immutable history and provenance on every asset. **Coming soon.**
Deployment engine that ships your signed artifact anywhere. **In private preview.**
***
## Community & Support
Get early access to Helix and Sign, plus a direct line to the founding team.
* [Try Viorant Hub free](https://viorant.ai)
* [Join the program](https://viorant.ai)
Questions, issues, or feedback are welcome.
* Start with [installing the Hub](/getting-started/installation)
* Browse the [core concepts](/concepts/agents)
* Check [troubleshooting](/troubleshooting) for common issues