FlatClaw, Private AI Platform
All use case spotlights
Operations AutomationCompliance & ApprovalsIT & MSP

Managed hosting operator

≈ $15M revenue (approx.) · 3,000+ hosted accounts · a 24/7 operations team

Host-panel and DNS operations agent

Account provisioning, DNS changes and site deployments through approval-gated connector tools, with each user's own panel credentials scoped to that user and every action in the audit log.

Organization
Managed hosting operator
System
A commercial hosting control panel with a broad API
Runs on
The operator's own tenancy
Scope
Operations agent with approval-gated actions
The situation

Where they started.

Routine hosting operations, account provisioning, DNS changes and deployments, were done by hand through the control panel, one screen at a time. The operator wanted the speed of an agent without the risk of an agent deleting a zone or an account on its own.

What FlatClaw does

What was built.

  • A first-party MCP connector for the control panel exposing the operations staff actually perform, with each user's own panel credentials scoped to that user.
  • Reads and routine changes run directly; destructive or exposing calls pause in the approval queue for a human.
  • Deploying a static site becomes a sentence: build, upload, extract, verify, with the agent reporting what it did.
  • Every action, approved or denied, recorded in the audit log.
Results

What changed.

  • Operations that took a sequence of screens take one request.
  • No unattended destructive action is possible by construction.
  • Credentials never shared between users or stored in a prompt.
  • The connector pattern reused for the operator's other systems.
Why private

Panel credentials and customer account data cannot pass through a third-party model. The connector holds them in a vault and the model only ever sees the request it composed.

Runs on

The operator's own tenancy

The stack

Components involved.

Hosting control-panel MCP connectorApproval enginePer-user credential vaultFlatClaw PortalPrivate inference
Your workflow

Have one like it?

Every spotlight started as a conversation about a process nobody liked doing, under a data-locality constraint.