Zoho CRM MCP Server and Zia Agents Guide 2026

Zoho CRM MCP connects AI assistants like Claude, ChatGPT and Cursor to Zoho CRM through Zoho-hosted servers. Here is what the four servers do, how to set them up, how to keep them secure, and when to use Zia Agents or Deluge instead.

Zoho CRM MCP lets you connect AI assistants such as Claude, ChatGPT and Cursor directly to your Zoho CRM, so they can query, create and update records in plain language. Zoho provides pre-built, Zoho-hosted MCP servers for CRM, and the wider Zoho MCP service is included at no extra cost for Zoho users; every call uses your existing CRM API credits and your own CRM permissions. Zia Agents is Zoho's own agent platform, where you build agents inside Zoho rather than in an outside AI tool.

This guide explains what Zoho CRM MCP and Zia Agents are, what each one can do, how to connect Claude, ChatGPT and Cursor step by step, how to keep it secure, and when a custom Deluge or API build is still the better choice.

Quick answer

Zoho CRM MCP = Zoho-hosted connectors that turn CRM actions into tools any MCP-compatible AI client can call. Four pre-built CRM servers exist: Data Insights (read-only), Data Operations, Module Customization and Workflow & Process Automation. Find them in Zoho CRM under Setup > Developer Hub > MCP for AI Agents, add the URL to your AI tool, and sign in with OAuth. Zoho MCP has no separate fee; it uses your Zoho API limits. Zia Agents are built in Zia Agent Studio; the platform is free and you pay only for model usage beyond the free token allowance. Start with the read-only server, a least-privilege user and human approval for any write.

What is Zoho CRM MCP?

MCP (Model Context Protocol) is an open standard that lets an AI assistant discover and call tools in another system. Zoho CRM MCP servers expose CRM capabilities as those tools. When you ask your assistant a question, it picks the right tool, Zoho CRM runs the action, and the result comes back into the conversation.

Zoho describes the CRM servers as pre-built, so you do not need to build, host or maintain any infrastructure. Zoho says they work with any MCP-compatible AI tool, and its documentation gives setup steps for Cursor, Claude Desktop, VS Code and Windsurf.

There are two related things with similar names:

Zoho CRM MCP servers: four ready-made servers published in the Zoho CRM developer documentation and found in your CRM's Developer Hub.
Zoho MCP (mcp.zoho.com): a broader console where you create your own MCP server, choose tools from Zoho apps and supported third-party services, and connect it to clients such as Claude.ai, ChatGPT, Cursor, Windsurf and VS Code. It also offers pre-configured servers, such as one for CRM data and metadata operations.

What can the Zoho CRM MCP servers do?

Zoho groups CRM capabilities into four servers. You can install more than one.

Data Insights

Read-only access to your CRM data. It can query records across standard and custom modules using COQL, filter, sort, group and paginate results, list all modules and read field schemas. Example prompt: "Show me all the deals closing this month."

Data Operations

Create, read, update and delete records, including bulk operations and related records, plus COQL search. This server can change and delete data, so treat it with care. Zoho's own example prompt deletes every deal stuck in a stage since March, which shows how much a single sentence can do.

Module Customization

Changes your data model: view module metadata, create custom modules, create custom fields, update field properties, and update, activate or deactivate page layouts.

Workflow & Process Automation

Create, update, list and reorder workflow rules, manage workflow task actions and read workflow configuration.

Zoho notes that API calls made through MCP consume API credits in the same way as standard API calls and are subject to your existing API credit limits.

What are Zia Agents and Zia Agent Studio?

Zia Agents is Zoho's platform for AI agents that run inside the Zoho ecosystem. Zia Agent Studio is the low-code builder: you give an agent a name, role and instructions, add documents to its knowledge base, attach tool groups so it can act (for example, update a record or add a task), and set guardrails as do's and don'ts. Zoho also lists observability and multi-agent features, and you can describe an agent in plain text and let Zia generate the configuration.

Agents can run on Zoho-hosted models or on external models. Zoho's model documentation lists bring-your-own-key support for vendors including OpenAI, Anthropic, Google Gemini and Cohere, with availability depending on your data center.

When were Zoho MCP and Zia Agents launched?

Zoho announced Zia Agents, Zia Agent Studio, the Agent Marketplace and the Zoho MCP server on 17 July 2025, alongside its Zia LLM. At that point they were rolling out to early-access customers, with general availability expected towards the end of 2025. Check the current pages linked below for the latest status in your data center.

How much do Zoho CRM MCP and Zia Agents cost?

Zoho MCP: Zoho's pricing page lists it at $0, with no separate subscription. It runs within the API limits of your existing Zoho editions; third-party services follow their own API policies.
Zia Agents: creating, deploying and managing agents is free. Zoho-hosted Standard tier models include 30 million free tokens per month, then $1.00 per million tokens; the Pro tier (GLM 5) includes 20 million, then $3.00 per million, and is available on request in the US and IN data centers. With your own API key, the model vendor bills you directly. Local taxes apply.

Prices are as published by Zoho at the time of writing and can change.

How do I connect Claude, ChatGPT or Cursor to Zoho CRM?

Before you start, you need a Zoho CRM account with the right permissions and an MCP-compatible AI tool.

Step 1: Choose the server

In Zoho CRM, go to Setup > Developer Hub > MCP for AI Agents. Pick the server you need. Start with Data Insights, because it is read-only. If you want a custom tool set, create a server in the Zoho MCP console instead and add only the tools you need.

Step 2: Connect Claude Desktop

Open Claude Desktop and go to Customize > Connectors.
Click the add icon and choose Add Custom Connector.
Enter a name and paste the server URL into the Remote MCP Server URL field.
Click Add, select the new connector and click Connect to sign in to your Zoho CRM organisation.

For Claude.ai on the web, Zoho's MCP guide uses Claude's settings: open Connectors, add a web connector with the MCP URL, then click Connect and authorise.

Step 3: Connect ChatGPT

Zoho's MCP implementation guide gives these steps for a Zoho MCP server:

In ChatGPT, go to Profile > Settings > Apps and click Create App.
Enter a name and description, and paste your MCP server URL.
Select OAuth as the authentication type and click Create.
On the Zoho MCP authorisation screen, click Allow.

Menu names in AI tools change often, and availability can depend on your ChatGPT plan, so check the current screens if anything looks different.

Step 4: Connect Cursor or VS Code

From the Zoho MCP console, choose Cursor in the Connect section and click Add to Cursor for automatic setup, then authorise. For VS Code, the CRM Developer Hub offers an Install MCP Server in VS Code button, or you can add the server URL to your mcp.json with type http and start the server to authenticate.

Step 5: Authenticate and test

The first time a tool runs, Zoho starts an OAuth sign-in in your browser. After that, the tools appear in your AI client. Test with read-only questions first, such as record counts or a list of open deals, and check the answers against CRM reports before you trust them.

How do I keep Zoho CRM MCP secure?

Zoho states that every action is scoped to your Zoho CRM user permissions, so the assistant can only do what your CRM role allows. Zoho MCP uses OAuth-based authorisation and Zoho says agent actions are governed by access controls and audit trails. Those controls only help if you set them up well:

Start read-only. Connect Data Insights first and add write servers only when there is a clear need.
Use a least-privilege user. Connect with a profile and role that sees only the modules and records the use case needs. Avoid connecting as an administrator.
Limit the tools. In the Zoho MCP console, add only the tools a workflow needs. Zoho's own guide recommends keeping the tool count under 300 so the client stays accurate; fewer is safer.
Keep a human in the loop. Require confirmation in your AI client before any create, update or delete action, and do not allow bulk deletes or schema and workflow changes from chat in production.
Test in a sandbox. Try write and customisation prompts on a sandbox or test organisation first.
Review audit logs. Check the CRM audit log and API usage regularly to see what the assistant changed and how many API credits it used.
Treat MCP URLs as access points. Share them only with the people who should use them, and disconnect clients you no longer use.

For a structured way to test an assistant before it touches live data, see our guide on how to evaluate AI agents.

MCP vs Zia Agents vs custom Deluge or API: which should you use?

OptionWhere it runsBest forCost modelMain watch-outZoho CRM MCPYour AI tool (Claude, ChatGPT, Cursor and others)Ad hoc questions, reporting and assisted updates by people already working in an AI assistantNo MCP fee; uses CRM API credits plus your AI tool subscriptionPowerful write and delete tools; needs tight permissions and confirmationZia AgentsInside Zoho, built in Zia Agent StudioRepeatable agent workflows inside Zoho apps, with knowledge bases and guardrailsPlatform free; pay for model tokens beyond the free allowance or via your own keyModel availability varies by data center; agents still need testingCustom Deluge or APIZoho functions, workflows or your own serviceFixed, rule-based logic that must run the same way every time, or complex integrationsDevelopment time plus API creditsNeeds build and maintenance effort; no natural-language flexibility unless you add it

Many teams use all three: MCP for exploration and assistance, Zia Agents for recurring agent tasks, and Deluge or the API for deterministic rules. If your CRM data is messy or you are still moving systems, sort that first; our CRM migration services guide covers how.

What can you use Zoho CRM MCP and Zia Agents for?

Pipeline questions: ask which deals close this month, which have no next activity, or how many leads came in last week.
Record updates: log a call, update a deal stage or create a follow-up task from a conversation, with confirmation first.
Data quality checks: find records with missing fields or likely duplicates, then review them before any change.
Admin help: list fields, layouts and workflow rules to document or audit your setup.
Recurring agent tasks: use Zia Agents for follow-ups, email or ticket replies and report preparation that run the same way each time.

What are the limits of Zoho CRM MCP?

API credits: every MCP call counts against your CRM API credit limits, so busy assistants can use up the allowance.
Permissions: the assistant cannot do more than the signed-in user, which is a protection but also a limit.
One organisation per session: each session is scoped to the organisation you signed in with; to switch, disconnect and sign in again, and do it deliberately so tests never hit production by mistake.
AI accuracy: the model chooses which tool to call and how. Wrong filters or vague prompts can return wrong answers or change the wrong records, which is why read-only first and human approval matter.

If you need the assistant connected in a more controlled way, Insyrge also builds LLM and CRM integrations, AI agents, Zoho CRM automation and CRM API integrations.

Sources

Frequently asked questions

Is Zoho CRM MCP free?

Zoho lists Zoho MCP at $0 with no separate subscription. Calls use the API credits in your Zoho CRM edition, and your AI tool may have its own subscription cost.

Do I need to host an MCP server myself?

No. The Zoho CRM MCP servers are pre-built and hosted by Zoho. You add the server URL to your AI tool and sign in.

Which AI tools work with Zoho CRM MCP?

Any MCP-compatible client. Zoho documents setup for Claude Desktop, Claude.ai, ChatGPT, Cursor, VS Code and Windsurf.

Can the AI see all my CRM data?

No. Actions are scoped to the permissions of the Zoho CRM user who signed in, so connect with a user that only sees what the use case needs.

Can the AI delete records?

Yes, through the Data Operations server, if the user's role allows it. Start with the read-only Data Insights server and require confirmation for any write.

What is the difference between Zoho MCP and Zia Agents?

Zoho MCP connects outside AI tools to Zoho apps. Zia Agents are agents you build and run inside Zoho with Zia Agent Studio.

Need help with Zoho CRM MCP and Zia Agents?

Insyrge will help you set it up end to end.

We set up Zoho CRM MCP and Zia Agents for you: the connection to Claude, ChatGPT or Cursor, permissions and least-privilege access, agent workflows, and testing before anything touches live data. Book a call with Insyrge.