Lead Response Engine
Inquiries sit unanswered for hours. By the time you reply, they've booked with whoever answered first.
$2,000setup, once
$600per month
Connected Operations Layer (MCP)
Your tools don't talk to each other.
Your inbox, your CRM, your calendar, your WhatsApp, your documents: every one is its own island, and you are the integration nobody's paying for.
$5,000 setup, then $2,500 a month. Fixed price, US dollars. Live in 21 business days from a signed scope and the deposit.
Free audit, no accounts, we change nothing.
Or ask a person: aimarketing@gonzo.co.in +91 7390990227 WhatsApp
Connected Operations Layer (MCP) is one of Gonzo Digital's nine fixed-price systems. It costs $5,000 to set up and $2,500 a month, in US dollars, and goes live 21 business days after a signed scope and the deposit. It is for businesses whose team wastes hours moving information between disconnected tools, ready to run one governed AI operator across all of them. Anything it sends on its own says it is from an AI.
What you get
- One AI operator connected across your whole stack (mailbox, calendar, CRM, documents and WhatsApp Business) through the Model Context Protocol, acting as a single surface instead of ten disconnected apps (today only a few read tools are built; most write connectors are not, and the router has never run live)
- Every irreversible action (sending, paying, publishing, deleting) stays human-approved by construction. The agent drafts and proposes; you approve. This is the safety that makes 'connect everything' sane
- WhatsApp handled through the official Business API within its real limits: own-account messaging and inbound replies, never the stranger-outreach automation that WhatsApp's policy does not allow and that can get accounts suspended
- Your accounts connect through the provider's own sign-in and you can revoke access at any time; where access tokens are held depends on the connection route and is confirmed in writing before go-live
- A full audit log of everything the agent read and every action it proposed, so you can prove what happened, not just trust it did
Who it is for
Businesses whose team wastes hours moving information between disconnected tools, ready to run one governed AI operator across all of them.
How it behaves
- Anything this system sends to a person on its own says it is an AI. There is no setting to turn that off.
- Fixed scope. The deliverables above are the scope, and the price does not move with the hours.
What this page does not show
Client results. This page describes what the system can do, not what it has done for someone. We will publish a result only when a real client has earned it and agreed to be named.
More detail
How it runs
- Connects your mail, calendar, CRM, documents and WhatsApp Business through the Model Context Protocol, each account through its own sign-in.
- Reads and drafts across all of them from one place.
- Anything irreversible waits for the approval of a named person on your side. That is by construction, not a setting.
Where it connects
Each row says what exists today and what has to be true before it runs for you. "Tested in the sandbox" means tested against simulated services.
| Connection | Status today | What going live needs |
|---|---|---|
| Microsoft 365 (mail and OneDrive) | Four read tools built and tested. | Your sign-in through Microsoft's own consent screen. Write actions are not built. |
| Google Drive | Two read tools built and tested. | Your sign-in through Google's own consent screen. |
| Gmail or Outlook, Google Calendar, HubSpot, other CRMs | Through an integration router that has never run live. No production tool map exists yet. | A router key, the tool map for your accounts and a first live test. Most write connectors are not built. |
| WhatsApp Business API (official) | Through the same router, never run live. | A WhatsApp Business API account. Own-account messaging and inbound replies only. |
| Gonzo's CRM store | Read and write executor built. | Used for the sandbox. Whether it fits your CRM is checked before you sign. |
| A durable store for the audit log and the approval queue | Built for SQLite, with a Postgres adapter. | Somewhere to keep it, ideally on an encrypted disk. |
What this does not do
- It is the least finished of the nine. The router has never run live, there is no connect-link flow yet, there are no inbound triggers, and every write connector except the CRM one is still to build.
- Every write is a proposal. That includes writes the underlying gateway would run at once. Nothing is sent, paid, published or deleted without an approval token from one of your named approvers.
- The approval screen itself does not exist yet. The code checks the approver against your list and refuses machine names, but authenticating the approver is the job of that screen.
- A credential is never accepted as a tool argument. A key or a value that looks like a password or token is refused. Accounts connect through the provider's own sign-in, and you can revoke them at any time.
- WhatsApp goes through the official API within its limits: your own account and inbound replies. It does not write first to people who have not opted in, because WhatsApp's policy allows contact only with people who opted in and says accounts that break it can be suspended.
- The proposals and their parameters are stored as plain text, so the store belongs on an encrypted disk. Email bodies are among them.
- One process is assumed. A failing shared dependency stops this agent for every client.
Doubts people raise about this one
Can I turn it off?
Yes, there is a kill switch. Every send goes through one checkpoint, and halting it stops the sending. There is also a circuit breaker that stops a failing action from repeating. A STOP from a customer is honored at once. The system ships switched off, and turning it on is one recorded step by a named person.
For the Connected Operations Layer, accounts connect through the provider's own sign-in and you can revoke access at any time; where access tokens are held depends on the connection route (an integration router is named in its stack) and is confirmed in writing before go-live.
What is not on file: who holds the switch for your system day to day, and the cancellation and minimum-term terms. A person will confirm both before you sign.
Can I see what it did?
Yes. Each action is appended to a tamper-evident log: who, what, a hash of the content, and the hash of the entry before it. Changing or deleting an entry breaks the chain and shows up when the chain is checked.
For the Connected Operations Layer the deliverable is: a full audit log of everything the agent read and every action it proposed.
A limit worth knowing: the chain proves nothing was changed afterwards. It does not prove what the system did was right.
Does it work with my CRM?
The single-purpose systems write to Gonzo's own CRM log, an activity record per contact. Whether that mirrors into your CRM is not on file for most of them.
The Connected Operations Layer is the one built to work inside your own tools. It connects through an integration router to your mailbox, calendar, CRM (HubSpot is named) and documents.
For any CRM not named here, a person will confirm whether and how it connects.
What does it connect to?
What exists today, by product. Everything below has been built and tested against simulated services only; nothing has run live for a client.
- Lead Response Engine: the official WhatsApp Business API and email are built. The web form and web chat adapters are not built yet. Your CRM: it writes to our own CRM log, and whether it mirrors into yours is checked per CRM.
- AI Receptionist: web chat and a calendar adapter (Google or Microsoft) are built against a sandbox calendar. The phone line and voice gateway are not built. Confirmations go by email.
- Client Onboarding Pipeline: an e-signature webhook, an intake form, email, WhatsApp (when a template is supplied) and a calendar are built in the sandbox. The calendar tools have never run live, and the store for onboarding state is not built.
- Dormant List Revival: email and the official WhatsApp Business API are built in the sandbox. It needs a separate sending domain and a persisted consent store.
- AI Outreach & Lead Qualification Engine: email and official WhatsApp (opted-in contacts only) are built in the sandbox. Apollo enrichment has never run live.
- Quote-to-Invoice Automation: an intake form, email and Stripe payment links (money in only) are built. Stripe has never been run live. Your accounting package is not connected: invoices live in the system's own book.
- E-commerce Support Assistant: an adapter for your storefront is written per platform. Email, WhatsApp and web chat are built in the sandbox.
- Connected Operations Layer: four Microsoft 365 read tools and two Google Drive read tools are built. Gmail or Outlook, Google Calendar, HubSpot and WhatsApp go through an integration router that has never run live, and most write connectors are not built.
Your exact version of a tool, a self-hosted system or a niche CRM is not on file. A person will check it before you sign.
Which AI does it use, and does my data go to another AI company?
The models are Hermes and DeepSeek, and the code allows no others. Which product uses which:
- AI Outreach & Lead Qualification Engine: our product list says it runs only on Hermes and DeepSeek.
- E-commerce Order, Returns & Product Support Assistant and Connected Operations Layer (MCP): the stack lists Hermes and DeepSeek.
- Every other product: a person will confirm in writing which models are used, because the product list does not say.
Rules and templates decide what is said. The model only helps with wording, and its wording is thrown away if it changes a figure or adds a claim.
Before text reaches a model, email addresses, phone numbers and money amounts are swapped for placeholders by pattern, and names are swapped where they have been supplied to the tool. An address, a name nobody supplied or an order number is not caught, so this lowers the risk and is not proof that no personal data remains.
Both model families are hosted. We have not yet settled whether hosted models may see redacted personal data, so we do not claim that your data never reaches another AI company. How personal data is handled is on the trust page. What the model providers' own terms say about training is not on file, so we do not state it. Nothing here has run live yet.
Who owns my data, my accounts and the workflow?
You own the workflow, the account, and the data from day one, not us. Credentials and documentation are handed to you in writing before we mark this live, not after. This is a gate in our own delivery process, not a promise.
That paragraph is in every proposal, for every product.
What we cannot answer yet: the contract wording on data return after you cancel, and the exact export format. A person will confirm both in writing before you sign.
Related
- Weighing the alternatives: Build it yourself with Zapier, Make or n8n, or buy it done, covering wiring your tools together yourself in a workflow tool.
- Managed Automation & Infrastructure Retainer: keeps it monitored once it is live.
- Lead Response Engine: a narrower system if inquiries are the whole problem.
- Prices in rupees are on the India price page. Your own numbers go in the calculator.
Other systems
See what you get →
Dormant List Revival
You have thousands of past customers and no system that ever talks to them again.
$1,500setup, once
$500per month
Client Onboarding Pipeline
A signed contract sits in an inbox for a week before anyone does anything with it.
$2,500setup, once
$700per month
Start with the free audit.
Send one web address. In about five business days you get a short written report. If a system fits, a fixed-price scope follows.
Free audit, no accounts, we change nothing.
Or ask a person: aimarketing@gonzo.co.in +91 7390990227 WhatsApp