Planet horizon lit by an orange sunrise, hero for Viorant AI deployment infrastructure for agents

Deploy an agent
from your terminal.

Deploy an agent
from your terminal.

vio is the Viorant CLI. It turns a folder into a .vio Space, deploys it to Viorant Cloud, and hands you a URL.

curl -fsSL https://dl.viorant.ai/cli/install.sh | sh

Install vio

Login

Or

download the Hub for macOS / Windows

Free for personal use. No credit card required.

vio CLI terminal demo: deploying an AI agent from a signed .vio file to Viorant Cloud
vio CLI terminal demo: deploying an AI agent from a signed .vio file to Viorant Cloud
BACKED BY BUILDERS
Supported by industry leaders, not gatekeepers.

01 Sign in

vio login

Google sign-in in your browser. One owner-only credential file on your machine.

02 Initialize

vio init

Turns the folder into a Viorant Space: agent.vio and its files, ready to edit. Works offline, no account needed.

03 Deploy

vio deploy agent.vio --target viorant_cloud

You get a URL and a run key, shown once. Runs on your Viorant credits, no provider key required.

See the full walkthrough →

✦ THE VIORANT MODEL

The artifact is the unit of

identity.

Meaningful AI objects keep distinct identity; .vio composes them into the complete system definition.

✓ Author locally with the Hub

✓ Test with built-in suites

✓ Govern with policy and approvals

✓ Prepare supported deployments with Helix

Explore the model →

Author locally
⬡ my-agent.vio
artifact: my-agent
version:1.2.0
type: agent

runtime:
 model: anthropic:claude-3-7
 tools:
 - web.search
 - code.interpreter

policy: policy.vio

tests:
 - unit
 - safety
 - regression
.vio
.vio
✓ Test & validate
Built-in suites ensure quality and safety.
✓ Govern
Policies and approvals baked in.
Supported targets
Cloud
On-prem
Edge / Device

✦ THE PROBLEM

Production AI has no durable source of truth.

The model lives in one dashboard, prompts in a repository, memory in another service, policies in documents, and deployment logic in provider-specific pipelines. The running system is assembled from state no single artifact records.

Every hand-off makes you answer the same questions again:

? which artifact is running in production?

? which prompt, model, and policy belong together?

? what changed after the last approved version?

? can we verify who signed this release?

? can this move to another framework or cloud?

? can anyone reproduce it from the source?

? unanswered at the next deployment▮

Fragmentation compounds across the lifecycle:

Drift

Local, staging, and production stop describing the same system.

Lock-in

Framework and cloud choices become permanent architecture.

Risk

Unsigned assets move through production without provenance.

Delay

Engineers rebuild deployment context for every environment.

Audit

Evidence is reconstructed after the system has already run.

✦ WITH AND WITHOUT VIORANT

Two architectures. The same teams.

Toggle between three separate stacks and one shared .vio definition that Trust proves and Helix prepares for supported targets.

Fragmented architecture

Three teams. Three stacks. No shared definition.

One-off runtime pathManual handoff
Viorant | Build an AI agent once, deploy it everywhereThe product, AI platform, and research teams each use a different framework, LLM provider, memory system, and runtime. Traffic moves through each one-off path while security policies are copied manually and no shared artifact records the full system. TEAMFRAMEWORKLLMMEMORYRUNTIME Product teamCUSTOMER AGENTLangGraphADAPTER SET AOpenAI LLMMODEL CONTRACT ARedis memorySCHEMA AAWSPIPELINE A AI platformINTERNAL COPILOTCrewAIADAPTER SET BAnthropic LLMMODEL CONTRACT BVector DBSCHEMA BGCPPIPELINE B Research teamEVALUATION AGENTLlamaIndexADAPTER SET COpen-weight LLMMODEL CONTRACT CLocal sessionsSCHEMA CPrivate VPCPIPELINE C 0SHAREDARTIFACTSSecurity + GRCTHE SAME POLICIES ARE COPIED INTO THREE FRAMEWORKS, THREE MEMORY SCHEMAS, AND THREE DEPLOYMENT PIPELINES.

Another framework does not

Another framework does not

solve fragmentation.

solve fragmentation.

AI needs a portable definition

AI needs a portable definition

independent of the runtime.

independent of the runtime.

The asset should outlive the editor, model provider, framework, cloud, and hardware chosen to execute it.

One definition across the AI deployment lifecycle.

Viorant keeps definition, provenance, transformation, deployment, and governance connected without sitting in the inference path.

Define

→

Sign

→

Verify

→

Transform

→

Deploy

→

Observe

✦ THE PLATFORM

Six products. One artifact lifecycle.

01

Hub

Available now

Create, test, and manage AI assets in one local-first workspace.

So you can keep the source and its dependencies visible before anything ships.

Explore Hub →

Viorant Hub icon, the local workspace for building AI agents

02

.Vio

Specification scope

Declare each AI asset as a portable, versioned artifact linked to its sub-files.

So you can review and move models, prompts, agents, skills, memory, connectors, and policies without hiding their relationships.

Explore .vio →

.vio icon, the portable AI system definition file format

03

Trust

Current scope

Fingerprint, sign, and verify artifacts with cryptographic manifests.

So you can prove origin, integrity, version, and approval before deployment.

Explore Trust →

Viorant Trust icon, AI artifact signing and verification

04

Helix

Supported targets

Transform signed .vio artifacts for supported frameworks and deployment targets.

So you can change the destination without redefining the asset around each provider.

Explore Helix →

Helix icon, the Viorant AI deployment engine

05

Viorant Cloud

Status not public

Use a managed deployment path while keeping the source definition portable.

So you can run through Viorant infrastructure without making it the permanent home of the asset.

Viorant Cloud icon, managed execution for AI agents

06

Sentinel

Status not public

Connect monitoring, guardrails, compliance, and evidence to the artifact lifecycle.

So you can carry governance into operation instead of adding it after deployment.

Sentinel icon, Viorant credits, billing and governance
✦ ROADMAP

Build the durable asset first.

Hub establishes the local source. Each later product adds portability, trust, deployment, or governance around the same .vio artifacts.

Now

Author

HubLive

Author, test, preview. Local-first, free.

Coming soon

Define and deploy

.vioScope published separately

One portable artifact.

HelixSupported targets

Prepare supported deployments.

Viorant Cloud

Managed runtime.

In development

Attest

TrustScope published separately

Prove origin, version, approval.

Next

Govern

Sentinel

Observe, audit, catch drift early.

Compress

Prompt and context compression.

The artifact remains yours as the surrounding platform expands.

✦ MODEL-AGNOSTIC & PROVIDER-NEUTRAL

Choose from supported models.

Declare model requirements explicitly and use providers supported by the current product and documentation.

Named providers·Explicit support·1 portable format
OpenAI
Anthropic
Mistral
xAI
Cohere
DeepSeek
gqGroq
Ollama
LMLM Studio
vLLMvLLM
✦ CONNECTORS & MCP

Connect through supported interfaces.

Declare connector requirements explicitly and use the current supported integrations and MCP interfaces.

Named connectors·MCP-native·1 protocol
Browse the connector catalog
Atlassian
Gmail
c7Context7
GitHub
Stripe
Cloudflare
grGranola
Sentry
Sequential
Thinking
HubSpot

✦ WHAT CHANGES

From runtime dependency to infrastructure you control.

Today

With Viorant

✕ Assets scattered across repositories and dashboards

✓ Readable .vio artifacts linked to their source files

✕ Definitions bound to one framework or provider

✓ Portable source transformed for each supported target

✕ Deployment rebuilt for every environment

✓ A consistent path from signed source to runtime package

✕ Production changes with no verifiable lineage

✓ Fingerprint, signature, version, and approval attached

✕ Governance reconstructed after deployment

✓ Policy and evidence connected across the lifecycle

✦ From the founder

The systems we are building do not think. They produce outputs that resemble thinking, often more useful than thinking, yet the mechanism is statistical, not human. The distance between what a model produces and what we can verify it understands is the epistemic gap, and it widens by the day.

The response to a gap one cannot close is not faith. It is structure: signed artifacts, declared agents, versioned policy, sovereignty by default. We govern what we cannot fully comprehend. We have always done so with people. We must now do so with their instruments.

That is the work. To

civilize intelligence.

Vivien Roggero

· Founder & CEO, Viorant

NO RIP AND REPLACE

Keep your stack.

Own the definition.

Viorant does not replace Git, your frameworks, models, or cloud. Hub is local-first, you bring your own keys, and .vio keeps the source portable. Your deployed runtime calls its models and tools directly. Viorant does not become a tollbooth.

git

supported frameworks

target adapters

supported clouds

named models

mcp

byok

local-first

QUESTIONS? WE’VE GOT ANSWERS.

Frequently asked questions.

01Is Viorant open source?+
.vio is designed as a neutral, portable specification. Review the current public repository, license, and released specification for its present status.
02How does Viorant ensure data sovereignty?+
Deployment and service choices determine data boundaries. Review the current security documentation for the exact product and release you use.
03Can I use my preferred models and infrastructure?+
Use the current support matrix to choose implemented models, interfaces, and deployment targets. The system definition remains separate from those choices.
04How is provenance and audit handled?+
Trust can fingerprint, sign, and verify stated artifact evidence. Verification proves only what the implemented evidence and policy check.
05How do updates and versioning work?+
Artifacts are designed for explicit versions and reviewable change. Current behavior depends on artifact type and product release.
06Can I self-host the platform?+
Customer-controlled deployment depends on the current supported target and its documented limitations. Contact Viorant with your required environment.
07What’s the pricing model?+
Free covers 3 deployed agents with 2,500 one-time Vio Credits. Starter is $15 a month for 5 agents and 10,000 credits a month; Pro is $50 a month for 10 agents and 60,000 credits a month. Plans are flat, with no hosting fee: deployed agents pay only for model calls, in Vio Credits.
08Is there an SLA for enterprise customers?+
Contact sales for current enterprise availability, support, security review, and contractual service terms.

Define the asset.

Keep control of where it runs.

Civilize intelligence.

Start locally with Viorant Hub, with 2,500 free credits included.

Download the Hub, free

macOS

Windows