Managed cloud
We run Spine for you, so there is no infrastructure to stand up. The fastest start, with each customer isolated as its own tenant.
Your cloud, your region, your models
Deployment
Run Spine where your regulator expects your data to be, on the models you choose, over the systems you already have.
Options
Four ways to run Spine.
Every option runs the same control plane with the same gates. What changes is who operates it and where.
In your region
Run in the cloud region your regulator expects, such as Google Cloud Mumbai for India, where Voice AI already runs today.
Your own cloud account
Spine ships as containers with its state in Postgres, so it is designed to run inside an account you control, such as Google Cloud or AWS.
Bring your own models
Point the model router at the models you already license, or at open weight models running in your region.
Regions
Spine maps regulation for the US, Canada, India and the UK and EU. Tell us where your customers are and we will propose the region and model set that fits.
Connections
How the control plane connects.
Products report up to Spine. Spine reaches your systems only through governed connectors.
Four products, one control plane
Voice AI, LiveAssist, NEQQO and Collections report to the same approvals, guardrails, kill switch and ledger. Turn on one product or all four.
Connectors for existing systems
One connector interface covers CRMs such as Salesforce, HubSpot and Dynamics 365, knowledge bases and your own services. Swapping a vendor is a registration change.
MCP tools
Import tools from any Model Context Protocol (MCP) server into Agent Studio. They sit behind the same allow list and write gate as every other tool.
Reads flow, writes wait
Anything not marked as a read is treated as a write, and every write waits for a named person. Connecting a system never gives an agent a free hand in it.
ConnectorsMCPModel routerTenant isolation
Onboarding
What an onboarding looks like.
One workflow first, live under gates, then the next. Each step is small enough to review.
Pick one workflow
Together we choose a first workflow and write down the rules it must follow, the systems it touches and who approves what.
Set up your tenant
Choose the region and models, connect the systems the workflow needs and name the people who approve writes.
Build and test
Build the agent in Agent Studio, dry run it, run evals and let it watch or suggest on real traffic before it acts.
Go live under gates
Raise the autonomy ceiling step by step. Approvals, the ledger and the Compliance Center carry the evidence from day one.
Questions
Common deployment questions.
No. Spine sits over the systems you have and reaches them through connectors. Your records stay where they are.
Yes. Each deployment is its own region with its own tenants and audit chains, so a new region is a new deployment rather than a migration.
Only people with the right role. A new agent version goes through the same checks and approvals, and from the Cockpit an operator can pause a product or lower its autonomy at any time with a reason on record.
Plan your first governed workflow.
Bring one workflow and your region. We will sketch the deployment, the connectors and the approvals it needs.