Zoho launched MCP support in mid 2025. It lets AI agents like Claude connect directly to Zoho CRM, Books, Desk, and 15+ other apps, reading data and taking action without manual input. The setup takes under an hour. The test we ran against 183 stalled leads produced a ranked re engagement list in minutes. This post covers the setup, the test, and where it falls short.
KEY TAKEAWAYS
Zoho MCP gives AI agents the ability to act inside Zoho apps, not just read from them.
The setup at mcp.zoho.com requires no coding. Basic configuration takes under an hour.
Custom servers work across multiple Zoho apps. Pre configured servers suit single function teams.
In a real test, Claude analysed 183 stalled CRM leads, scored each one, and identified which were worth re engaging, without a human touching the data.
Write operations carry real risk. Test in a sandbox before connecting to live production data.
The agent is only as good as the Zoho data it connects to. Messy CRM = messy output.
The Problem Zoho MCP Is Solving
Asana’s Anatomy of Work report found that 60% of work time goes to what they call ‘work about work’, status updates, searching for information, and context switching between disconnected tools. The figure is widely cited. What it describes is easy to observe: a sales rep who knows which deals need attention still has to open four apps, locate the right records, and manually execute the follow up. The intelligence and the action remain separated.
Zoho MCP closes that gap. It is not a new interface or a chatbot layer. It is a connectivity standard that lets an AI agent operate inside your Zoho apps, reading live data, applying judgement, and executing tasks, without a human routing each step.
This post covers what Zoho MCP is, how to connect it to Claude, and what actually happened when we ran it against a real Zoho CRM account with 183 stalled leads.
What Is Zoho MCP?
MCP stands for Model Context Protocol. It is an open standard developed by Anthropic, the company behind Claude, to give AI agents a standardised way to communicate with business software. Before MCP, connecting an AI model to a business application required custom API integration for every combination of model and tool. MCP provides a shared communication layer so any MCP compatible AI model can connect to any MCP compatible application using the same protocol.
The most useful analogy is HTTP. HTTP made it possible for any browser to load any website, regardless of who built either. MCP does the same thing for AI agents and business applications: it removes the bespoke integration requirement.
Zoho built its own MCP implementation at mcp.zoho.com. The platform turns every Zoho app into what Zoho calls an ‘agent ready endpoint.’ Any MCP compatible AI, Claude, GPT, Cursor, or others, can connect to Zoho CRM, Books, Desk, Projects, and 15+ other apps to read data and execute actions, within the user’s existing permission structure. There is no vendor lockin on the AI side. The same Zoho MCP setup works regardless of which AI model you use.
IMPORTANT DISTINCTION
Zoho MCP is not a chatbot builder, a prompt template library, or another drag and drop automation tool. It is the execution layer between an AI agent and your business operations.
A chatbot replies. An automation tool follows preset rules. An AI agent connected via MCP interprets your intent, locates the right data, and carries out a multi step task, all from a single prompt. The difference is the absence of a predefined script.
Where Zoho MCP Fits in Zoho’s AI Story
Zoho has been running AI features since the early days of Zia, its builtin intelligence layer. By the time Zoho published its major AI update in July 2025, Zia had over 80 algorithms running across the suite, lead scoring, anomaly detection, sales forecasting, email sentiment analysis, call transcription, best time to contact suggestions.
The gap was the last mile. Zia would surface the insight. It would flag the lead, predict the deal, identify the anomaly. Then a human would navigate to the right app, find the right record, and do something with that insight. The intelligence and the action were separated by a manual step.
On 17 July 2025, Zoho announced three things simultaneously: Zia LLM (its own large language model), Zoho AI Agents (purpose built agents for specific business tasks), and Zoho MCP (the connectivity layer that lets those agents act across the Zoho suite). The three components form a stack where Zia provides the intelligence, Zoho AI Agents provide the autonomy, and Zoho MCP provides the connectivity.
What took Zoho until 2025 to ship is worth acknowledging. The MCP standard only became viable for production use in 2024 and 2025 as tooling matured. Zoho was building on a protocol that did not exist at production scale any earlier.
Layer
What It Does
Example
Zia Artificial Intelligence
Surfaces insights, predictions, recommendations
Flags a high priority lead based on behaviour patterns
Zoho AI Agents
Takes autonomous action on business tasks
Follows up with the flagged lead automatically
Zoho MCP
Connects the agent to the right Zoho app to act
Updates the CRM record and sends the follow up email
Zoho Apps
Where the action actually executes
Lead updated, email sent, task created in Projects
What Zoho MCP Can Actually Do
The table below shows the cross app capability set as it stands. The more instructive illustration is the 183 lead test further into this post.
Task
App
What the Agent Does
Lead prioritisation
Zoho CRM
Pulls records, reads interaction history, scores leads by stated criteria
Invoice creation
Zoho Books
Creates, formats, and sends client invoices end to end from a single prompt
Support ticket routing
Zoho Desk
Opens ticket, assigns to queue, pulls customer interaction history as context
Task creation
Zoho Projects
Creates tasks with deadlines and assignees across active projects
Campaign scheduling
Zoho Campaigns
Configures and schedules campaigns by target segment
Team notifications
Zoho Cliq
Sends contextual updates based on CRM triggers or status changes
Expense logging
Zoho Expense
Logs and categorises expenses from a description prompt
Three use cases are worth walking through because the workflow change is not obvious from a list
Lead analysis without running a report
Salesforce’s State of Sales report found that sales reps spend roughly 34% of their time actually selling. The rest goes to administrative work, data entry, and finding information across tools. With Zoho MCP connecting Claude to Zoho CRM, that information retrieval step becomes a direct query:
Show me all deals in negotiation that have had no activity in the last seven days, and flag the ones where the contact is a C suite decision maker.
The agent pulls the records, applies the filters, and returns the list. No report builder. No manual pipeline review to find out the best deals are going cold.
Invoice creation without opening Zoho Books
For small teams where one person handles both sales and billing, the context switch between Zoho CRM and Zoho Books is a genuine time drain. A single prompt handles the full sequence:
Create and send an invoice to Acme Corp for the Q1 consulting project. $4,500, due in 30 days.
The agent creates the invoice in Zoho Books, formats it, and sends it. What typically takes five to seven minutes of menu navigation takes one prompt.
Support ticket routing with customer context
With Zoho Desk connected via MCP, a support team lead can open and route a ticket while pulling customer history in the same action:
Open a high priority ticket for Brynka Inc., assign it to the enterprise support queue, and pull their last three interactions as context.
The agent does not just create the ticket. It brings the customer history along so the assigned agent has context before opening the record. Three separate actions, one prompt.
How to Connect Zoho MCP to Claude: Step by Step
The basic configuration requires no coding. The process from first login to a working connection takes 15–30 minutes for a custom server, less for a pre configured one.
Step 1: Go to mcp.zoho.com
You will see two paths: pre configured servers (pre built setups for specific CRM functions) and a custom server option. For businesses working across multiple Zoho apps, start with a custom server, it gives full control over which apps and actions the agent can access.
Fig. 1, Zoho MCP homepage at mcp.zoho.com showing pre configured server options and custom server creation button
Step 2: Create your MCP server
Click ‘Create MCP server.’ Name it something functional, ‘zoho crm claude’ works fine. The name is for your reference only and has no effect on how the server functions.
Fig. 2, Create MCP server dialog box with server name input field in the Zoho MCP dashboard
Step 3: Add tools to your server
Your server starts empty. Click ‘Add Tools’ to open the Zoho product catalogue: CRM, Workdrive, Cliq, Qntrl, Projects, Mail, Bigin, Desk, Books, Billing, Inventory, Invoice, Expense, Payroll, SDP On Demand, Apptics, and more. By default, Zoho tools require authorisation on the first tool call. You can pre configure a specific Zoho account in the Connection tab if you prefer.
Fig. 3, Zoho product catalogue showing all available apps that can be added as tools to a custom MCP server
Step 4: Select Zoho CRM and choose your actions
Select Zoho CRM. You will see 169 available actions, get records, search leads, create contacts, update deals, retrieve webhooks, pull workflow rules, and more. All 169 are selected by default. Narrow them to what your workflow actually needs. A smaller permission surface reduces the risk of the agent doing something unexpected. Click ‘Add Now’ when ready.
Fig. 4, Zoho CRM tool selected in the MCP catalogue with 169 available API actions listedFig. 5, Zoho CRM actions selection screen showing all 169 operations available to the AI agent
SETUP NOTE
Start with 12–15 actions rather than all 169. The core operations, get records, search leads, create contacts, retrieve related records, cover most workflows. Add more once the initial setup is tested and stable.
Step 5: Review your linked tools
Your linked tools now appear in the server dashboard: Get Records, Get Roles, Get All Tasks, Get Profile, Get Updated Field Details, Get Related Records, Get Webhooks Usage Reports, and the others you selected. Add more apps using ‘Add More Tools’ in the top right.
Fig. 6, Linked tools review screen in Zoho MCP dashboard showing all CRM functions connected to the server
Step 6: Get your MCP URL
Go to the ‘Connect’ tab. You will find your unique MCP URL endpoint with an API key embedded. Copy this URL. The same page shows setup instructions for Cursor, Windsurf, and VS Code if you use those clients instead.
Fig. 7, Connect tab in Zoho MCP server dashboard showing the unique MCP URL endpoint with embedded API key
Step 7: Open Claude and go to Settings
In Claude.ai, click your profile icon at the bottom left of the sidebar and select ‘Settings.’
Fig. 8, Claude.ai settings menu accessed via profile icon showing Settings, Language, and Get Help options
Step 8: Navigate to Connectors
In the Settings sidebar, click ‘Connectors.’ You will see built in connections for Google Drive, Gmail, Google Calendar, and GitHub. This is where you add Zoho MCP as a custom connector.
Fig. 9, Claude Connectors settings tab showing built in connections for Google Drive, Gmail, Google Calendar, and GitHub
Step 9: Add a custom connector
Click ‘Add custom connector.’ Name it ‘ZohoMCP’ and paste in the URL from Step 6. Claude will display a trust confirmation noting that the connector has not been verified by Anthropic and that you are responsible for all actions taken. Review this, confirm, and click ‘Add.’
Fig. 10, Add custom connector dialog in Claude settings with name and integration URL input fieldsFig. 11, Claude trust confirmation prompt for the Zoho MCP custom connector before adding to the workspace
NOTE FOR TEAMS
Only a Claude organisation admin can add a custom connector for team wide access. Once added at org level, it becomes available to everyone in the Claude organisation. Individual users on a personal plan can add it independently.
Step 10: You are connected
ZohoMCP appears in your Connectors list, labelled ‘CUSTOM.’ Start a new conversation in Claude and it will have access to your Zoho CRM data and actions, and any other Zoho apps you added as tools.
Fig. 12, Claude Connectors list showing ZohoMCP successfully added and labelled as CUSTOM connector
Pre Configured or Custom: Which Server Type to Use
The pre configured servers at mcp.zoho.com are not a lesser option. They are faster to set up and well tested for single function workflows. If your use case fits into one CRM function, lead management, deal tracking, contact operations, the pre configured server is the faster path.
Custom servers are the right choice when a workflow spans multiple Zoho apps in the same sequence: pull a deal from CRM, generate an invoice in Books, open a follow up task in Projects, all from one prompt.
Pre Configured Server
Custom Server
Setup time
Under 5 minutes
10–15 minutes
Flexibility
Limited to preset tools
Full control over apps and actions
Best for
Single function workflows
Multi app workflows and agencies
Technical skill
None
Minimal, no coding required
Starting point for
First time Zoho MCP users
Power users and Zoho heavy businesses
We Tested It: 183 Stalled Leads, Ranked With One Prompt
This is not a product demo scenario. It is a real test run on a real Zoho CRM account with real sales data.
The setup: a sales team had 183 leads in the ‘Not Responding’ stage, all tagged with Project Technology set to Zoho. Conversations had happened, emails exchanged, calls made, and then things went quiet. Manually reviewing 183 interaction histories to decide who to call first would have taken the better part of a working day.
Instead, Claude, connected to Zoho CRM via Zoho MCP, was given one prompt:
Give me the emails of the leads sitting in the Not Responding stage where Project Technology is set to Zoho. Then analyse the conversation and interaction history for each one and score them out of 10 for likelihood of re engagement.
Claude did not return a filtered list. It built its own scoring methodology on the fly, applied it across all 183 records, and returned a tiered priority table with scores, reasoning, and risk flags.
Fig. 13, Claude’s initial response to the 183 stalled leads prompt, pulling records from Zoho CRM via MCPFig. 14, Claude’s scored lead prioritisation table showing 183 leads ranked by re engagement likelihood with scoring rationale
The scoring methodology Claude used
Scoring Criterion
What Claude Looked At
Recency of last meaningful interaction
Highest weight. A conversation from six months ago scores higher than one from two years ago with more exchanges.
Explicitness of their response
Did they say 'send me a proposal' or something vague? Clear buying signals ranked up.
Seniority of contact
Owners and C suite ranked above staff contacts for the same company.
Company fit for Zoho services
Industry, company size, and technology profile assessed against typical Zoho deployment patterns.
Conversation depth
Leads who asked specific Zoho questions ranked higher than those who gave generic interest.
Whether their original question was answered
Unanswered questions flagged as high re engagement potential.
Claude also surfaced something the team had not asked for: the large majority of the 183 leads were cold outreach attempts from 2022 and 2023 with no email addresses on file. It recommended these be moved to a backlog review queue rather than active re engagement.
Fig. 15, Claude’s flagged output identifying leads with no email addresses on file and recommending backlog triage
One thing the output also showed: two leads were flagged as high priority despite having bounced email addresses. The scoring logic did not catch that because it was not built into the prompt. You still have to check the output. The agent surfaces patterns faster than a human, it does not replace the judgement call at the end.
The difference between this and a CRM report is worth stating directly. A report filters data. An AI agent connected via Zoho MCP reads conversations, applies scoring criteria, builds a methodology, and produces a prioritised action plan, at the speed of a query, not a working day.
What This Changes, and for Whom
The practical impact looks different depending on where the bottleneck sits in your operation.
Small business operators running lean teams
The administrative overhead of CRM hygiene, invoicing follow ups, and support ticket routing is a real cost in single person or two person operations. A properly configured Zoho MCP setup shifts the data retrieval and cross app coordination to the agent, freeing the people in the operation to focus on work that requires human judgment.
Sales managers with active pipelines
The 183 lead test reflects a Tuesday morning pattern, not a special project. Stale pipeline reviews, cold lead triage, and deal stage analysis are recurring tasks that eat into time that would otherwise go to actual selling. With Zoho MCP connected, those analyses become on demand queries rather than scheduled reports that arrive after the deals have already gone cold.
McKinsey’s 2024 State of AI report found that companies adopting AI assisted workflows reported productivity improvements three to four times higher than those using traditional processes. The report’s own caveat: the gains concentrate in teams that have connected their AI tools to real operational data, not just used AI for content generation. Zoho MCP is the infrastructure that makes the former possible.
Support team leads managing ticket volume
Zoho Desk connected via MCP makes ticket status, resolution history, and agent workload queryable in natural language. For teams managing high volumes, the more significant change is the context bundling: an agent that opens a ticket and pulls customer history simultaneously removes the step that currently forces support agents to start cold on every new case.
What to Know Before You Go All In
It is early stage infrastructure
Zoho MCP launched in 2025 and is actively being developed. The pre configured servers are solid for the use cases they are designed for. Complex custom workflows spanning multiple apps simultaneously will require iteration before they run reliably. The 183 lead test was read heavy and performed well. Write heavy workflows carry more variability and need staged testing.
Write operations require a sandbox first
Read operations, pulling records, searching leads, retrieving history, are low risk. Write operations, creating records, sending emails, updating deals, carry real risk if the agent misreads a field or a multi step workflow fails partway through. Test against non critical data before connecting to live production.
Scope your permissions deliberately
The AI agent operates within whatever Zoho permissions the connected account holds. Connecting an admin account gives the agent admin access to everything in Zoho. Create a dedicated service account with role based permissions scoped to what the agent actually needs. Treat it the same way you would treat a new employee: access to what the role requires, not the master key.
Not all 169 CRM actions are equally stable
Core operations, get records, search leads, create contacts, retrieve related records, work reliably. More complex operations involving bulk updates, webhook configurations, and advanced module relationships behave less consistently. Start narrow and expand as you confirm each action works in your environment.
The output quality depends on the data quality
The 183 lead analysis worked because the underlying CRM data was structured and populated. An agent pointed at a messy CRM, missing fields, inconsistent lead stage definitions, empty interaction history, will return messy output. Zoho MCP amplifies whatever is already in the system. It does not clean it.
Setting up Zoho MCP for your specific operation
The technical connection is the straightforward part. Getting it to work well for your CRM field structure, support workflows, invoicing patterns, and lead stage definitions is where the configuration work sits. Clixlogix has implemented Zoho environments since before MCP existed. We now work with clients on MCP connected agent setups, from making the underlying Zoho configuration agent ready, to scoped permission architecture, to testing workflows against live data safely.
Zoho MCP is Zoho’s implementation of the Model Context Protocol, an open standard developed by Anthropic. It lets AI agents connect to Zoho apps, read live business data, and execute actions directly within those apps without manual input. Available at mcp.zoho.com.
Zoho MCP is available to Zoho users, but specific pricing and plan inclusion vary by subscription level. Check mcp.zoho.com directly for current pricing.
Zoho MCP officially supports Claude (Anthropic), Cursor, Windsurf, and VS Code based agents. Because it uses the open MCP standard, other MCP compatible models can also connect, which is the main advantage over proprietary integrations.
The current list includes Zoho CRM, Books, Desk, Projects, Mail, Workdrive, Cliq, Inventory, Invoice, Billing, Expense, Payroll, Bigin, Qntrl, SDP On Demand, Apptics, and more. The platform continues to expand.
Zia is Zoho’s built in AI assistant, predictive analytics, lead scoring, anomaly detection, sentiment analysis across the Zoho suite. Zoho MCP and Zoho AI Agents are the action layer built on top of Zia’s intelligence foundation. Zia identifies what should happen. Zoho MCP is what makes it happen.
The basic setup at mcp.zoho.com requires no coding. Select your server type, choose apps and actions, copy the MCP URL, paste it into your AI client. Advanced configurations with custom server logic or scoped permission architecture benefit from someone familiar with API concepts, but the core workflow is accessible to non technical users.
Zoho’s automation tools, Workflow Rules, Blueprints, Deluge scripts, follow preset, rule based logic. They execute the same steps in the same order every time. Zoho MCP gives an AI agent the ability to interpret natural language, decide which tools to use, and execute multi step tasks without a predefined script. Automation is rigid and condition based. MCP is flexible and context aware.
Written By
Chief Executive Officer
As CEO of Clixlogix, Pushker helps companies turn messy operations into scalable systems with mobile apps, Zoho, and AI agents. He writes about growth, automation, and the playbooks that actually work.
Google's biggest Maps overhaul in over a decade just changed how local businesses get discovered. Here is what local SEO professionals and business owners need...
