Connect Claude or ChatGPT to Your Zoho Apps: What’s Safe and What to Ask First

Chances are some of your team are already pasting Zoho data into an AI chat window to summarise a deal, draft a reply or check an invoice. It works, and it is also a quiet way for customer data to leave your systems with no record of it.

Zoho MCP offers a more controlled route. Instead of copying data out, an AI assistant such as Claude or ChatGPT is connected to your Zoho apps and works inside them, under rules you set. That is a real step forward. It also raises questions every business should answer before switching it on. Here is how it works, what Zoho says keeps it safe, where the gaps are, and an eight-point checklist.

What Zoho MCP is, in plain English

MCP, the Model Context Protocol, is an open standard that lets AI assistants use business software. Zoho MCP lets you build a server that exposes chosen tools, actions and data from your Zoho apps to an AI assistant. Zoho describes it as model-agnostic and compatible with Claude, GPT and other agents. It lists apps including CRM, Desk, Projects, WorkDrive, Books and Billing, and sample workflows such as updating deals, replying to tickets, generating invoices and logging expenses. Signing up is free.

Setup follows a simple pattern. You create a server in Zoho MCP, choose which tools to enable, and copy the server URL. Then you add that URL as a connector in your AI assistant. For Claude, Zoho’s help pages say you need to be an organisation admin on a Team plan or higher. For ChatGPT, you need a Plus, Pro or Team plan with Developer Mode switched on.

What Zoho says keeps it safe

You choose the tools. Zoho says the assistant does not automatically get every action. It can only use the tools you enable.
Sign-in runs through OAuth. You approve access on a Zoho consent screen that lists the permissions being requested. Zoho says agents act under user-level permissions and can only do what that user is allowed to do.
Actions are logged. In the example we read, the Zoho MCP Logs page records each tool call with its status, the user ID and an execution ID, and you can filter by date, status or tool.

These are the right foundations. They are not the whole job.

Where the gaps are

The server URL is a key. Zoho’s guidance is to treat it like a password, because anyone who has it can use your enabled tools. Regenerating it disconnects every client.
Your data goes to an AI provider. Zoho’s Analytics MCP page warns that the data gathered for your request is shared with the AI model to produce the answer. The same page says that server is still in a development phase.
Some details are not published. The pages we read do not say how long logs are kept, whether they capture what was asked or only what was done, or what the usage limits are. Ask Zoho before you rely on them.
AI can misread instructions. Zoho’s own advice is to review AI actions before treating them as final, and to test in a sandbox first.

The governance checklist: eight things to settle first

Start with the job, not the tool. Write down one or two things you want the assistant to do, such as “show me deals that have gone quiet for 14 days”. If you cannot name the job, you are not ready to connect it.
Choose who it connects as. The assistant can reach what the connected user can reach. Use a user whose role matches the job, not a Super Admin.
Enable the fewest tools you can. Start with read-only tools. Add write tools one at a time. Leave delete switched off.
Decide what needs a human. Anything that leaves the building, such as an invoice or a customer email, and anything that changes money or deletes records, should be reviewed first. Emails and tickets can also contain text written by outsiders, so do not assume everything the assistant reads is trustworthy.
Protect the server URL. Decide who holds it, store it like a password, and regenerate it when someone leaves.
Know where the data goes. Check your AI provider’s plan and data terms. List which Zoho apps hold sensitive data, such as financials or health records, and keep those out until you are sure.
Own the logs. Name one person who reviews the audit log on a schedule, and confirm with Zoho what it records and for how long.
Test, then trust. Use Zoho’s playground or a sandbox, begin with reversible tasks, and keep a review period before anything runs unattended.

Where we would start

Start by reading, not writing. Questions like “which invoices are overdue?” or “which tickets are waiting longest?” save time and cannot change a record. Once your team trusts the answers, add one low-risk write at a time, such as drafting a follow-up note for a person to approve.

One more tip. If you find yourself typing the same request every morning, that is not a chat. It is a process, and it probably belongs in a Zoho Flow or a Zia agent where it runs the same way every time.

Not sure where your Zoho is ready for AI?

This is the diagnose-before-we-prescribe work. We will look at your Zoho roles and permissions, help you choose a first use case, and set up a connection with the least access it needs, so your team gets the benefit without the worry. Book a free consultation at consultjc.com or email info@consultjc.com.

People first. Systems that scale.

Catalyst 3.0: Zoho’s New Platform for Building AI Agents, Not Just Apps

A question we hear more and more from growing businesses: “Can we build something custom, and can it live on Zoho’s own infrastructure?”

On September 2, 2026, Zoho gave that question a bigger answer. Catalyst 3.0, the latest version of its developer cloud platform, is positioned as an “agent-ready” full-stack cloud, built so AI coding assistants can take an app from a prompt to production without a developer stitching infrastructure together by hand. Here is what that means in plain English, what to ask before you rely on it, and when we would, and would not, point you at it.

What Catalyst is, in plain English

Catalyst is Zoho’s cloud platform for building and running custom applications: the hosting, the back end, the data, the scheduled jobs and the rest of the plumbing, all inside Zoho’s own cloud. Zoho’s argument for version 3.0 is that writing code is no longer the slow part. It says more than 72% of developers now use AI daily to write code, but provisioning, connecting and operating the cloud services behind it is still a bottleneck. Catalyst 3.0 is meant to remove that step.

“Agent-ready” means the platform is designed to be driven by AI coding assistants such as Claude Code and Codex, not just by people clicking through a console. Zoho’s Catalyst page lists support for Claude, Cursor, Antigravity, Codex and GitHub Copilot.

What’s new in 3.0

Agent Skills. Packaged knowledge that teaches an AI assistant how Catalyst works, so it generates code that fits the platform. They are listed on the Claude Code and Codex marketplaces.
MCP support. Lets a developer run Catalyst actions, such as deployment and configuration, from inside their coding assistant instead of the Catalyst console.
A non-interactive CLI. Runs tasks end to end without anyone answering prompts along the way. The Skill also decides whether each request goes through the CLI or MCP, rather than leaving that choice to the AI model.
Full-stack services. Front-end hosting, serverless functions, compute and data management, with hooks into Zoho applications. A few open-source language models are hosted on Zoho’s infrastructure for apps that need AI features.
Open-source pieces. The Skills repository, the SDK, the CLI and Slyte, a lightweight JavaScript framework, are all open source, which means your developers can inspect them.
Regional hosting. Zoho says you can host in the US or EU for compliance-sensitive apps, or in its India data centres for data sovereignty.

Governance: what Zoho says, and what to ask

“Built-in governance” is the headline claim, and for AI agents that can change live systems it is the part that matters most. Zoho says apps inherit the security layer of its cloud. Catalyst’s own page adds that agents log in with their own configurable permission set, that destructive commands are disabled in non-interactive mode, that there is an audit trail covering application, platform and MCP tool-call logs, and that changes after launch are versioned, attributable and reversible.

Those are the right ideas. One independent analyst, IBRS, was more cautious in a September 28 review. It found the announcement “stronger on promise than on proof”, with little detail on security architecture, audit specifics or compliance certifications. So before you commission anything, ask for evidence:

Does every agent action run under its own traceable identity, and can we see the logs?
Which compliance certifications apply to the region we would host in?
Which third-party MCP servers will the agents connect to, and who has vetted them?

What it costs

Pricing is pay-as-you-go, with a per-unit cost for each feature, a monthly free tier, and free credits for six months (listed as the equivalent of about ₹15,000, or $250 on Catalyst’s site). The Console shows usage and lets you set budget alerts and spending ceilings. Turn those on from day one, because AI agents can spend quickly.

When we would point you at Catalyst, and when we would not

Our approach is native first. Zoho CRM, Creator, Books, Flow and Deluge already cover most of what growing businesses need, and every line of custom code is something you have to own, maintain and explain to the next person. We try the simplest path before we complicate anything.

Catalyst starts to make sense when a real requirement sits beyond what the native tools do well. Think of a customer-facing portal with its own logic, a heavy integration, or a product feature that needs its own back end. Go in with open eyes about the trade-offs the analysts flag too: generated applications suit prototypes, dashboards and simple data apps better than complex, business-critical systems, and AI-written code builds maintenance debt from day one.

Five questions to ask before commissioning a custom build

Why can’t native Zoho do this? If you cannot answer that clearly, start there.
Who owns it in two years? If your key developer resigned tomorrow, could someone else pick it up?
Who reviews AI-written code? Speed is only a gain if someone checks what the agent produced.
Where does the data live? Pick the region deliberately and confirm it with your compliance lead.
What is the smallest pilot? Start with one small team and one non-critical application, and measure delivery time, quality and the effort spent on governance.

Thinking about a custom build on Zoho?

We will start by diagnosing what you actually need, tell you honestly whether native Zoho can do it, and, if it cannot, help you scope something simple that your team can own. Book a free consultation at consultjc.com or email info@consultjc.com.

People first. Systems that scale.

Schedule a Free Consult

Tag: MCP