Zoho Analytics MCP server: AI agents get a semantic layer
Zoho Analytics adds an MCP server, a Claude plug-in and a semantic layer for AI agents, at unchanged pricing from $24/month. Define your metrics first.
Zoho Analytics now has an MCP server, so AI agents can query your business data directly. Zoho announced on October 28 that its BI product is becoming an "agentic analytics" platform: an MCP server in remote and local versions, a new CLI, a Claude plug-in, and a built-in semantic layer that tells an agent what "revenue" or "active customer" means at your company. Pricing did not change. The Basic plan still starts at $24 a month, billed annually, and now includes LLM credits.
What actually happened
Per Zoho's Business Wire release:
- MCP server: remote and local options, with tools for data analysis, data preparation and visualization. Custom agents can also call external MCP servers.
- Claude plug-in and CLI: you can work with Zoho Analytics from inside Claude, or script it from the command line. The release gives no setup details for either.
- Semantic and ontology layers: a business glossary (metrics, synonyms, instructions) plus a model of how your entities relate. Zoho says definitions are portable through the Apache Ossie spec. The semantic model and ontology arrive as previews over the next four to eight weeks, not GA.
- Governance: admins can see which agents ran, who used them, what they accessed and how they answered. A sandbox lets you test agents and roll back.
- Connectors: more than 30 new ones, including PayPal, Adobe Commerce, Twilio, Notion, Intercom and Supabase, on top of 500-plus.
- Pricing: unchanged. Basic is $24/month billed annually for 2 users, 0.5 million rows and 10K Zoho LLM credits. On-premises customers get the update in Q1 2027.
Why a semantic layer for AI agents matters for your business
Ask an agent "what was our margin last month?" against raw tables and it will guess. It picks a column, ignores refunds, counts a test order. The answer looks confident and is wrong. That is the real failure mode of "chat with your data," and an MCP server alone does not fix it.
The semantic layer is the part to care about. It is where you write down, once, that net revenue excludes refunds and shipping, and that a customer is "active" after two orders in 90 days. Every agent, dashboard and Claude chat then reads the same definition.
Our advice, whether you use Zoho or not:
- Write your top ten metric definitions before you connect any agent. Plain English is fine. If your team argues about a definition, that argument is the work.
- Start agents read-only. Use the audit view to check what they touched in week one.
- Wait for GA on the semantic model before you rely on it for numbers that go to a lender or an investor.
- Keep definitions portable. Zoho's Ossie claim is a good sign; test an export before you commit.
Key takeaways
- Zoho Analytics adds a remote and local MCP server, a CLI and a Claude plug-in
- Semantic model and ontology are previews rolling out over four to eight weeks
- Pricing is unchanged: Basic from $24/month billed annually, now with 10K LLM credits
- An agent without metric definitions guesses; write yours down first
- Start agents read-only and check the audit log
Want an agent that answers margin questions correctly? We define your metrics, wire them to your data and build agents that read from one source of truth, on systems you own. See how we build it or tell us what you need to know weekly.
Sources: Zoho press release via Business Wire, Enterprise Times.
- #zoho-analytics
- #mcp
- #semantic-layer
- #ai-agents
- #business-intelligence
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
RingCX puts AI and human agents on one QA rubric
RingCentral RingCX now supervises, scores and forecasts AI agents next to human agents, with two-way handoff. Score your AI agent like staff.
Read itGoogle TPUs on Oracle Cloud in 2027: What It Means for AI Costs
Oracle will offer Google TPU 8t and TPU 8i chips on OCI starting in 2027, with no pricing yet. Plan AI costs on hardware you can rent today.
Read it