For AI agents

Run your business from your AI agent.

Connect Claude, ChatGPT, Claude Code, Cursor or any agent that speaks MCP. It can look up customers, answer leads, book appointments, draft invoices and campaigns, and more, with exactly the permissions you have in the app, and it asks you before anything reaches a customer or moves money.

Two addresses.

For your business

https://ultimatebusinessmachines.com/mcp

Your agent signs in with your account and asks you what it may do: only look, or look and make changes. Or use a personal token from Settings > AI agents.

Just looking

https://ultimatebusinessmachines.com/mcp/public

No account needed. Your agent can learn what we do, tailor a plan to your business with the Solutions guide, show what that plan costs, open the demo, and give you a sign-up link.

Connect in a minute.

  • Claude

    Settings > Connectors > Add custom connector, and paste https://ultimatebusinessmachines.com/mcp. Claude opens our sign-in, then asks what it may do.

  • ChatGPT

    Settings > Apps & Connectors > Advanced: turn on developer mode, then create a connector with the same address.

  • Claude Code

    claude mcp add --transport http ubm https://ultimatebusinessmachines.com/mcp

    Signs in through your browser, or add --header "Authorization: Bearer <token>" with a personal token.

  • Cursor, VS Code and others

    Add an HTTP MCP server with the address, and a personal token as Authorization: Bearer <token>. Apps that show only so many tools can use a token with the compact list.

What your agent can do.

  • Anything you can do in the app, for the services on your plan: your customer list, leads, bookings, jobs, invoices, reviews, your website and more.
  • Ask before it sends: emails and campaigns to customers, review requests, posts.
  • Write your texts for you. They still go from your own phone, unless you connected your own texting number.
  • Add or remove services, after showing you exactly what it costs.

What stays yours.

  • Signing up, choosing your password and agreeing to the terms.
  • Paying for your plan, and asking for your money back.
  • Saying a customer agreed to marketing texts, or vouching for an address or a signature for them.
  • Who gets your new-lead alerts, and the phone and email your customers reach you at.
  • Approving an agent, and disconnecting it. An agent can never give itself more access.

Safe by design.

  • Your permissions, no more.

    A team member's agent can do what that team member can do. Services you're kept out of, and administrator-only actions, stay that way.

  • Look only, if you like.

    Choose "look only" when you connect, and the agent can read but never change or send anything.

  • Disconnect any time.

    Settings > AI agents lists every agent, when it was last used, and a Disconnect button that works at once. Administrators see the whole team's.

  • Important changes on the record.

    Your activity log names the agent and connection behind the actions it records, such as access and plan changes, opt-outs and some deletions. It doesn't record every change, and Settings > AI agents shows when each agent was last used.

  • Your data, your agent.

    What your agent reads goes to the AI company behind it, under your agreement with them. We only give it what you could see yourself.

  • Built on the open standard.

    The Model Context Protocol (versions 2024-11-05 to 2026-07-28), with OAuth 2.1 sign-in: PKCE, short-lived tokens, and no passwords shared with the agent.

Not a customer yet?

Let your agent find your plan.

Point it at https://ultimatebusinessmachines.com/mcp/public, or tailor a plan here yourself.

Tailor a plan to your business Create your account