Deployment and trust

Your data stays in your platform, and you can verify it

VaultSpeed runs the control plane as a managed single tenant service and holds metadata, never your data. This page lists what runs where, who is accountable, and the evidence you can verify today.

Book a working session Trust center

Where things run

Three ways to deploy

VaultSpeed holds metadata, and the data never leaves your platform. Deployment covers a range, from fully managed to almost everything inside your perimeter.

Fully managed

SaaS

VaultSpeed runs the control plane, the agents and the LLM gateway as a managed single tenant service. The generated code lands on your platform.

Control plane managed, sessions yours

Hybrid

Agent sessions run in ephemeral sandboxes on your infrastructure, against your LLM gateway and your approved LLMs, with spend capped at the platform level.

Nothing leaves

Inside your perimeter

Almost everything runs inside your network, for organizations where nothing may leave it. The control plane directs the work, and the work runs next to your data.

Metadata, not data

Holds metadata, not your data

VaultSpeed works from the data model, the source structures, the mappings and the lineage. The data plane stays with you: the generated code is delivered into your CI/CD process and runs on your platform without VaultSpeed in the loop. There is no runtime dependence and no licensed engine required to keep the code executing.

Bring your own LLM. The LLMs you have vetted are used as they are, with no markup on inference and no token resale, so routing and spend stay with you.

What VaultSpeed holds
  • The data model and its definitions
  • Source structures, keys and mappings
  • The lineage of every transformation
  • The generated code, versioned in Git
What stays with you
  • Your data, on your platform
  • The runtime: the code runs without VaultSpeed
  • Your LLM contract and your token spend
  • Your Git, your CI/CD and your credentials
Agents under control

Nothing ships without review

Agents propose, people and reviewing agents approve. Every proposal is reviewed and versioned in Git before it reaches your platform, and every session runs inside limits your operators set.

Reviewed before it ships

Every agent proposal goes through Git and code review. Engineers and reviewing agents check each change before anything reaches your warehouse.
Reviewed before it ships

Ephemeral sandboxes

Every agent session runs in an ephemeral sandbox on Kubernetes, created for that run and removed afterward. Only what is written back to the context store persists.
Ephemeral sandboxes

One LLM gateway

LLM access runs through MLflow, which keeps the layer LLM agnostic, self hosted LLMs included, so inference runs where you decide.
One LLM gateway

Spend caps

Token and compute spend is capped at the platform level. Your operators set the limits, and no agent loop runs past them.
Spend caps
How review works on the platformTrace and reproduce a reported number
Inference choice

Bring your own inference

A reviewer checks three things here: where prompts go, which LLMs are approved, and whose bill the tokens land on. The gateway points at one of three places, and the answer holds for all three.

Full inference options on the Platform page

LLM gatewayone endpoint for every agent, pointed at one of three places
SnowflakeCortex REST API

Inference inside your Snowflake account

The gateway calls Cortex in the account and region you already run. Consumption lands on your Snowflake bill, under the roles and budgets you already govern.

  • No data or prompt leaves the Snowflake boundary
  • LLMs Snowflake has approved for your account
  • Spend on the contract you already have
DatabricksModel Serving

Inference inside your Databricks workspace

The gateway calls Model Serving endpoints in your workspace, governed by Unity Catalog. Genie works from the same semantic views the generator produces.

  • Inference stays inside the workspace
  • Unity Catalog permissions apply
  • Spend on your Databricks commit
Amazon Bedrockplatform neutral

When you run more than one platform

When your data is not on one platform, the gateway points at Bedrock. Two ways to run it.

Your BedrockYour AWS account, your region, your IAM policies and logging
Our BedrockProvisioned by VaultSpeed in your subscription's region, ready from the first session
  • Prompts and responses are never used to train models

Which LLM you use is a configuration, not an architecture decision. The context store, the skills and the review flow do not change when the inference does.

Customer story · Wealth management

We knew if we could get the hard part right first, everything else would fall into place. VaultSpeed made that possible.
One of Australia's largest wealth managers built a unified data foundation for scale, compliance and AIRead the story

Certifications and evidence

What is verifiable today

ISO/IEC 27001:2022 certified

VaultSpeed holds ISO/IEC 27001:2022 certification. The certificate and the full control set are published at trust.vaultspeed.com.
ISO/IEC 27001:2022 certified

Data residency per deployment

The control plane runs in the region of your subscription. The data plane runs in your own cloud account, so residency follows your deployment.
Data residency per deployment

Single sign-on and roles

Sign-in through your identity provider with SAML. Roles for operators, builders and reviewers.
Single sign-on and roles

Who is accountable

Your integrator keeps delivery accountability, and the review rules define who can ship what: agents propose, your reviewers approve, and every change is tracked and reversible in Git.

How we work with your partners

Open the trust center

Talk to us

Start with what a reviewer can check

Share this page with your security or governance team, then bring their questions to a working session with your own cloud setup as the reference.