For developers and AI builders
Sergio is a customer operations platform for residential cleaning companies, with an MCP server for AI agents.
Give your agent the business's synced booking and customer history (BookingKoala, Launch27, ConvertLabs and others), connected conversations, tasks and procedures, and the tools to act on them with permissions and approval controls.
Build lead follow-ups, VA morning plans and owner reports using the operational data already in Sergio.
Docs: Quickstart · Tool reference · ChatGPT and Claude setup
| Choose Sergio when… | Consider another approach when… |
|---|---|
| You're building for a residential cleaning business. | You need a general-purpose backend for unrelated industries. |
| The business uses Sergio or is willing to onboard. | You need access without the business adopting Sergio. |
| You need customer, booking and conversation context together. | You only need a standalone database or a single integration. |
| Your workflow fits Sergio's exposed actions and approval model. | You need actions Sergio doesn't expose. Check the tool reference first. |
Sergio's MCP lets your agent start with a cleaning business's operational context and available actions already connected. You focus on what your agent should do.
| What you'd build from scratch | What Sergio provides through its MCP |
|---|---|
| Connect booking, phone and email systems | One interface to the systems Sergio already connects |
| Match customers across those systems | A shared customer record with conversation and booking context |
| Translate raw data into operational decisions | Cleaning-specific context: recurring versus one-time clients, upcoming bookings, follow-ups |
| Give the agent ways to act | The actions Sergio exposes, such as creating tasks or campaigns |
| Maintain those connections | Sergio handles the underlying integration work |
For example, "Find customers overdue for a follow-up and prepare the next step" becomes an agent workflow instead of an integration project.
Clients, leads, cleaners, tasks, tags and lists are already modeled and connected. You start on the part that makes your product different.
No reverse engineering or brittle syncs of your own. Sergio pulls BookingKoala every night and reads its notification emails as they arrive; Launch27 and ConvertLabs sync nightly, read-only. Customer and job history is already in Sergio, ready to read and query.
Sergio creates and tracks tasks from what happens in the business, and clears them when the work is done. Your agent plugs into it instead of rebuilding to-dos and reminders.
Automated texts, emails and review requests already work, so you don't have to rebuild outbound.
Any client supporting remote MCP over Streamable HTTP, including Claude, ChatGPT, Cursor, Muse, Grok and custom agents. OAuth for AI clients, an agent key for scripts.
Built inside cleaning companies running it every day, so the edge cases come from real days, not guesses.
Contacts, history, notes, tags and lists. Search, read and edit.
Upcoming and past bookings with the customer and the team assigned. Read only.
Texts, emails and calls in one thread per customer. Draft replies for review.
Staff, their weekly hours, time off and scheduling rules. Read and update.
Build a campaign, choose its audience, write its messages and read the results.
Create, assign, snooze and complete the team's work.
The company's own procedures, plus training modules and quiz questions you can write and assign.
A cursor feed of new texts, emails, calls and booking changes. Poll it, then acknowledge what you handled. At least once; kept 7 days.
idempotency_key. A retry with the same key returns the original result instead of acting twice; the same key with different inputs is refused.Retry-After.https://sergio.serviche.com/api/agent/mcp to your agent or AI client. It is a remote server over Streamable HTTP, listed in the official MCP Registry as com.serviche/sergio.The worry that stops most owners from saying yes to an AI build is a text nobody meant to send. Sergio gives the owner the controls, so it's a feature you can sell.
create_draft writes a reply that waits in Sergio for the team to review.held: true and the reason, never silently dropped.Watch the events feed for new inquiries, read the customer's history, and draft the first reply and the follow-ups in the business's voice from its handbook. The office reviews and sends.
A weekly read on new leads and how fast each got a reply, cancellations and their reasons, and what each campaign brought in, written in plain language.
Each morning, read the day's schedule, open conversations and overdue tasks, and give the virtual assistant a short ordered plan as tasks in Sergio.
Turn the company's procedures into training modules and quiz questions, and assign them to new hires.
A remote MCP server that gives an AI agent a residential cleaning company's customers, bookings, conversations, team, tasks, campaigns and handbook, plus the actions Sergio exposes. The server address is https://sergio.serviche.com/api/agent/mcp and it is listed in the official MCP Registry as com.serviche/sergio.
Any client that supports remote MCP servers over Streamable HTTP, including Claude, ChatGPT, Cursor, Muse, Grok and custom agents.
Two ways. An AI client like Claude or ChatGPT signs in through OAuth, and the business owner or an admin approves it in Sergio. A custom agent uses an agent key that the owner or an admin creates under Add agent; Sergio shows it once, so store it as a secret. Either way your agent gets its own seat on the team with a role that sets what it can read and change, and the owner can revoke a key, pause the seat or remove it at any time.
Booking platforms like BookingKoala, Launch27, ConvertLabs and others for bookings and customer history, the business's phone line for texts and calls, and email. Your agent reads them through one interface instead of one integration each.
Drafts never send. A send takes two calls: the first returns the exact text, recipient and a confirm token and sends nothing; the second, with the token, sends. The business owner can switch Sending off, which holds every send, and opt-outs and quiet hours always apply.
Sergio pulls BookingKoala every night (between 1 and 4 AM local time) and reads its notification emails as they arrive. Launch27 and ConvertLabs sync nightly and read-only. Texts, emails and calls on the business line arrive as they happen.
Through a cursor feed, not a webhook: subscribe once, poll events_list, fetch details with read, then events_ack. Delivery is at least once, so dedupe on event_id. Events are kept 7 days.
Building and testing on our TidyHome demo business is free. A real business needs a Sergio subscription, which includes MCP access with no separate agent-seat fee; AI use beyond what the plan includes is billed separately. Each client business has its own subscription.
Yes. We connect your agent to TidyHome, our demo cleaning business with thousands of made-up customers, so you can build against real-shaped data with no real people in it. Access is by invitation for now.
Connect the Sergio MCP and try it on our demo business, TidyHome. Tell us what you're building and we'll give your agent a seat. Building for several clients? Say so; we're working out how we partner with builders.