Your AI analyst just got a major upgrade
More business teams are doing their thinking inside Claude and ChatGPT than ever before. Research, planning, analysis, content: it's all happening inside LLM platforms now.
But the moment someone needs an answer grounded in actual enterprise data, the workflow breaks. They leave the AI, open the BI tool, run the query, copy the result back. Context lost, momentum killed.
That's the problem we set out to solve when we launched ThoughtSpot's Agentic MCP Server back in July. Since then, Spotter, the most trusted agent for enterprise analytics,has gone through multiple major upgrades, and this release brings the full Spotter 3 experience to Claude, ChatGPT, and any MCP-compatible platform or custom agent you build.
What started as the ability to ask data questions and build Liveboards now includes deep research, root-cause analysis, code execution, and AI-generated summaries. Same governed analytics, grounded in your semantic layer and security model, with significantly more depth.
There are two ways to put this to work: plug-and-play access inside LLM platforms like Claude and ChatGPT, and embedding the Spotter agent as the analytics engine inside your own custom applications. We'll walk through both.<br><br>For a broader primer on MCP and why it matters for analytics, check out the Executive’s Guide to MCP.
What happens when you ask a question
Ask the Spotter agent a question in Claude or ChatGPT today, and what you get back is closer to what a senior analyst would hand you: not just the answer, but the reasoning behind it, the context around it, and the follow-up questions you should be asking next.
Here's what that looks like in practice. You ask about declining margins in a product line. The Spotter agent:
Runs multi-step deep research, breaking the question into sub-analyses, identifying contributing factors, and surfacing related patterns across your data
Performs root-cause analysis to explain why margins are declining, not just that they are, and returns diagnostic insights with AI-generated summaries in plain language
Executes code directly when questions go beyond standard queries, whether that's statistical modeling, custom calculations, or scenario analysis
Shows its reasoning at every step, so you can verify the logic before you act on the result
One more thing worth noting: you no longer need to know which data source to point your question at. The Spotter agent automatically selects the right model based on your query, which means business users can ask questions without knowing anything about the schema underneath.
Governed data without tool switching
For data leaders, the value here is governance. When business teams ask analytical questions inside Claude or ChatGPT, those queries are going through ThoughtSpot's semantic layer. Your business definitions, security rules, and access controls apply automatically, regardless of which AI platform the question came from. That's a significant shift: your teams work wherever they're most productive, and the data governance you've built doesn't get bypassed in the process.
For business users, the value is simpler: don't switch tools. If you're already working in Claude or ChatGPT and you hit a data question, just ask it.
Here's what that looks like across a few different teams:
Product. A PM investigating onboarding drop-off opens Claude and asks, "Where are users falling out of the activation flow?" The Spotter agent surfaces usage data right there. The PM digs deeper with follow-ups, and when they've found what they need, they can produce a Liveboard of the insights without ever leaving the conversation
Marketing. A lifecycle marketer asks ChatGPT, "Which customers have increased usage but haven't upgraded plans?" ThoughtSpot surfaces the segment. From there, the marketer asks ChatGPT to draft a targeted upsell email for that group. Analysis and action, same window.
Sales. A sales leader prepping for a new vertical asks Claude, "What do our highest-converting deals in healthcare have in common?" ThoughtSpot returns the patterns. The sales leader then asks Claude to turn those findings into a one-pager for the team. Here's how to connect the Spotter agent to your platform.
Ship agentic analytics without ever building a data agent
If you're building a customer-facing application, an internal tool, or a custom AI agent, there's a good chance your users will need to ask questions about data. And building the analytics layer to support that is a project in itself: natural language understanding, query generation, data governance, visualization. Most teams spend months on it before they ship anything.
The Spotter MCP Server gives developers a shortcut. Instead of building a data agent from the ground up, you plug the Spotter agent into your application as the analytics skill. Your users ask questions in natural language, and the Spotter agent handles the reasoning, query execution, and response generation behind the scenes. Your development team stays focused on what makes your product unique.
What makes this different from a basic API integration is the depth of what you're plugging in. This is the same Spotter agent with the same deep research, root-cause analysis, code execution, and AI-generated summaries covered in the previous section. Your users aren't getting a simplified version; they're getting the full analytical experience, embedded directly in your product, with ThoughtSpot's semantic layer and security model governing every response.
Get started
The Spotter 3 MCP Server is available now in early access for ThoughtSpot Analytics and ThoughtSpot Embedded enterprise customers.
Already a ThoughtSpot customer? Setup is a single URL. Connect the Spotter agent to Claude or ChatGPT and start asking questions against your governed data in minutes.
New to ThoughtSpot? Schedule a demo to see the Spotter agent in action.
