Montebelle
HomeHow It WorksThe OSCase StudiesTechnologyPricingAboutBlogContact Book a Call

How it works

Six steps from first call
to a live agent.

We provision the system. You bring the business. Your agent learns the rest from your first real conversations.

The flow at a glance

Onboarding is a guided product flow.

Not a DIY install. Not a webinar. A real onboarding path with clear ownership at every step.

01

Server

We provision your environment.

02

Runtime

Agent runtime installed and pinned.

03

Ops Center

Monitoring + control plane online.

04

Cognitive OS

Memory, verification, persona boot.

05

Channels

You connect your messaging apps.

06

First conversation

Agent learns from real exchanges.

In detail

What happens in each step.

01
Day 1

Server provisioning

Montebelle stands up your dedicated environment. Isolated by tenant. Audit logging on by default. Hardened to baseline before anything else touches it.

You don't pick a host, a region, a runtime version, or a model. The system is pinned to known-good defaults that we operate ourselves.

Provisioned by Montebelle~30 minutes
02
Day 1

Agent runtime install

The Montebelle agent runtime is installed and pinned to a verified release. The runtime is the same build we run on every fleet, including ours.

Upgrades happen on Montebelle's schedule, not yours. You don't get paged at 3am because a model provider changed an API.

Provisioned by MontebellePinned release
03
Day 1–2

Ops Center comes online

Your monitoring and control plane comes online. Health, conversations, approvals, audit, channel status. All in one place. Branded if your tier supports it.

The ops center is also where your authentication connections live. You'll use it to link the messaging accounts your agent operates from.

Run by MontebelleBranded on Enterprise+
04
Day 2

Cognitive OS boot

Memory, verification, and persona layers boot with your defaults. The agent knows it's an agent, knows the rules of the operation, and knows what it doesn't know yet.

This is the layer that makes verification gates real: before any irreversible action, the agent checks that its information is current and that the action is reversible.

Run by MontebelleMemory + verification + persona
05
Day 2–4

You connect your channels

Through the ops center, you connect the messaging apps your agent will operate on. WhatsApp Business, Signal, Slack, Discord, Telegram, iMessage, SMS, webchat. Your call.

Auth lives in your name. You can revoke at any time. Montebelle never holds your messaging account credentials directly.

Owned by youAuth stays in your name
06
Day 4+

First live conversation

This is where the agent actually becomes yours. Not by ingesting a 40-page playbook. Not by filling out forms. By talking to you about the work that's in front of you this week, on the channel you already use.

Memory is time-aware: the conversations you have in week one weight heavier than anything Montebelle pre-loaded. Your voice, your judgment calls, your "always do this / never do this" patterns become the agent's defaults, not a generic operator's. By the end of the first week the agent is handling the first class of work end-to-end with you in approvals; by week four most of those approvals have loosened to spot-checks.

Owned by youBootstrapped in conversationTime-aware memory

Responsibilities

Clear lines from day one.

No ambiguity about who runs what. Montebelle operates the system. You own the business.

Montebelle runs

The operational layer

So you never have to think about it.

  • Server provisioning, hardening, and uptime
  • Agent runtime, model selection, and upgrades
  • Ops center, monitoring, and incident response
  • Channel adapters and webhook reliability
  • Memory lifecycle, verification gates, fleet QA
  • Security patches and dependency hygiene
You own

Your business and accounts

Your auth, your context, your final say.

  • Your messaging accounts and channel auth
  • Business context, voice, and approval rules
  • Final decisions on irreversible actions
  • Your data, exportable on request
  • Brand and customer relationships
  • What the agent is allowed to do without asking

Launch timeline

What the first month looks like.

A typical SMB deployment. Personal is faster, Enterprise has more validation cycles.

Day 0 · Discovery call

15-minute call to map the operation

What channels, what processes, what would the agent do this week if it were already live? We pick a deployment tier together.

Day 1–2 · Provisioning

Server, runtime, ops center, cognitive OS

You receive ops center credentials and a kick-off doc. We're online and waiting before you connect channels.

Day 2–4 · Channel link-up

You connect your messaging accounts

Guided through the ops center. Auth is yours. Channels go live one at a time as you verify each.

Week 1 · Bootstrap conversations

Your agent learns the business

You talk to it like an operator joining the team. It asks questions when it should. Approvals stay tight.

Week 2–4 · Calibration

Approval rules loosen as confidence builds

What the agent is allowed to do without asking expands as you mark patterns "always do this." We tune memory lifecycle.

Day 30 · Operating

The agent is part of the operation

You're not approving every move. You're spot-checking. Most days the agent just runs, and you read the brief.

Get started

Want to see this for your operation?

Book a 15-minute call. We'll map your channels, pick a tier, and walk through what week one looks like.