Ask once, and get
an answer from
your whole stack.
Connect your PSA, RMM, docs and security tools to Claude, Claude Code, Cursor or any MCP client through one endpoint. You decide, person by person, who can look at what and who is allowed to change it. Every call is logged either way.
The console on the rightbelow is a scripted walkthrough on demo data. Nothing is called. Pick a role, pick a question, then approve or deny the change yourself. Switch to read only and watch the same request get refused.
integrations
a human decision
logged
You decide who
can do what,
person by person.
An admin sets the rules per person and per tool. Finance can read Pax8 invoices and nothing else. A first line tech can look at devices and propose a reboot, but only a lead can let it run.
The check happens twice: when the request is made, and again at the moment it runs. This table is wired to the console above, so whatever role you picked there is the one being shown.
Connect.
Expose.
Approve.
Set it up once. After that nobody has to hop between consoles to answer a simple question, whether they are on the service desk, in ops or doing the billing.
Add your tools once
Add each tool once. Its credentials go into your workspace vault, encrypted with a key that belongs to your org and nobody else's.
Set who can do what
Give each person a role. Claude and Claude Code sign in with your LayerOps account, so there is no key to paste. Clients that cannot sign in interactively, like Cursor, use a key you mint in Connect.
Ask anything, approve the rest
Questions are answered on the spot. If the AI wants to change something, it lands in your approvals list with the exact request it intends to send.
Twenty minutes
of clicking, or
one question.
Take a question like “which clients have a domain controller offline and a P1 open?”. Today that is three consoles and a couple of exports. Here it is one question, and the answer tells you where each part came from.
18 tools.
One endpoint.
18 native integrations across PSA, RMM, security, documentation, networking, monitoring, backup, distribution, automation and Microsoft 365. If yours is not here, give us its OpenAPI spec and we will build the tools from it.
Nothing changes unless
someone on your team
says so.
Questions are free to run. Anything that writes waits for a person with the right role, unless an owner or admin has written an allow rule for exactly that write. That rule lives in our server, so no prompt can talk its way around it.
Your keys stay in the vault
A KMS root key wraps your org key, and your org key encrypts every credential. Your AI only ever sees the result of a call.
Every write waits for a person
A proposed change is stored with the method, endpoint and payload it wants to send. If nobody acts on it, it expires after 24 hours.
Roles checked twice
Owner, admin, operator and read only, set per integration and checked again at the moment an action runs.
No made up endpoints
Calls are matched against a catalogue of real endpoints from each vendor's spec, so a client cannot invent one.
One price per workspace.
Bring as many people as you like.
For solo operators running their own stack.
- 3 integration connections
- 5,000 tool calls / month
- 4 built in roles
- 14 day log retention
For a whole team in one workspace: service desk, ops, finance, procurement.
- Unlimited team members
- 10 integration connections
- 50,000 tool calls / month
- Custom roles and policies
For orgs with compliance and SLA requirements.
- Unlimited integrations
- 250,000 tool calls / month
- 365 day log retention
- Audit trail CSV export
Connect your stack
in an afternoon.
Fourteen days free and no card needed. Connect a couple of tools, set who can do what, and see what your team asks it first.