The ConverSight MCP Server connects any MCP-compatible AI assistant directly to your organization's own ConverSight data, so every team member can ask questions in plain English, right inside the assistant they already use, and get answers pulled straight from live data — no SQL, no dashboards, no exports. MCP, short for Model Context Protocol, is an open standard for connecting AI assistants to outside systems; ConverSight speaks it natively, so setup takes minutes per person and works the same way across supported clients.
How MCP Server Helps
- Ask questions in plain English and get answers pulled from live ConverSight data, without writing SQL or building a dashboard first.
- Explore what's available before asking about the numbers — list datasets, describe a dataset's structure, or list the tables it publishes.
- Turn a query result into a chart, or validate and run a pipeline, without leaving the conversation.
- Connect the AI assistant your team already uses instead of adopting a new one — Claude Desktop and ChatGPT are both supported today.
- Give business users a self-service option for the ad hoc questions that would otherwise become a reporting ticket.
- Keep every connection scoped to the signed-in person's own ConverSight permissions, so nothing it returns exceeds what that person could already see.
MCP Server Features
| Feature | What It Means |
|---|
| Connect any AI assistant | Any MCP-compatible client can connect — this isn't a Claude-only or ChatGPT-only integration. |
| Row-level security (RLS) | Every connection runs under the signed-in person's own ConverSight account, so existing RLS applies exactly as it does anywhere else in ConverSight. |
| Tiered tool permissions | Administrators set each tool to run freely, ask first, or stay off entirely — from read-only exploration up to pipeline writes. |
| Cross-client support | The same connector and permission model work identically whether a team standardizes on Claude, ChatGPT, or both at once. |
| Domain-bound endpoints | Each connection is bound to the organization's own ConverSight domain, so it can only ever reach that organization's data. |
| Self-registering OAuth | Clients register themselves automatically (RFC 7591) on first connect — no client ID or secret to issue by hand. |
| Curated dataset scope | Only the datasets and streams a team has curated for MCP are exposed, never raw underlying storage. |
| Revocable access | Access can be reviewed or withdrawn at any time from the client's own connector settings. |
Use Cases
- Ad hoc reporting: "Total sales by each quarter" or "Show me last quarter's order lines by region."
- Data exploration: "Describe supply_chain_v2 and show me a few sample rows" before building out a question.
- Visual answers: "Show that as a bar chart by quarter," plotting a query result without switching tools.
- Operational pipelines: "Validate this pipeline before I save it" or "Run the weekly replenishment pipeline."
Get Started
| Client | Guide |
|---|
| Claude Desktop | Add ConverSight as a custom connector in Claude Desktop. |
| ChatGPT | Add ConverSight as an MCP App in ChatGPT's Developer mode. |