SyncRange can act as a Model Context Protocol (MCP) server, so assistants like ChatGPT, Claude and Gemini can read your connected data and answer questions in plain conversation — no export or spreadsheet required.
Pick the guide that matches how you want to connect. Each page walks through the UI (or config) for that product.
| Client | Best for | Auth |
|---|---|---|
| ChatGPT | Asking questions in the ChatGPT web / desktop UI | OAuth (recommended) |
| Claude | Asking questions in Claude.ai (and Claude Desktop connectors) | OAuth (recommended) |
| Gemini | Gemini Enterprise chat UI, or Gemini CLI | OAuth (Enterprise) or bearer token (CLI) |
| CLI & local clients | Cursor, Claude Desktop config, Gemini CLI, curl | Manual bearer token |
| This connection | It does not |
|---|---|
| Lets your assistant read summary data (e.g. "what were my Shopify sales last month?") from sources you choose | Let your assistant write, modify, or delete anything in your store, ad accounts, or Google properties |
| Returns aggregated, bucketed data sized to fit comfortably in a chat response | Dump your full raw order/customer history row-by-row through the chat |
| Applies your plan's usual data limits — each active assistant connection counts as one exporter, and rows read count toward your monthly rows limit | Bypass your plan's exporter or rows limits |
AI Assistants are built for quick analysis in conversation — not as a replacement for a warehouse or BI stack.
| A good fit | Better elsewhere |
|---|---|
| Ad-hoc questions (“how did Meta spend look last week?”) | Scanning years of raw order or customer rows |
| Quick checks while you work (sales, ads, traffic summaries) | Heavy joins, custom SQL, or multi-source modelling at scale |
| Exploring a date range that fits a chat-sized summary | Ongoing large historical archives and team-wide reporting |
For large data needs, treat SyncRange’s AI connection as the conversational layer on top of summaries — not the place to pull everything. Export into a data warehouse such as BigQuery first, then use AI (or BI tools) against that warehouse when you need deep history, row-level detail, or complex analysis.
All clients talk to the same SyncRange MCP endpoint:
https://mcp.syncrange.com/mcpThe same URL is shown on your dashboard under Connections → AI Assistants.
Once connected, ask a normal question — e.g. "what were my Shopify sales last month?" You don't need to paste ids yourself. If you've connected more than one account from the same source, refer to it by name; if your assistant isn't sure which one you mean, it'll ask you to clarify. See How your assistant finds the right account in the data reference.
The AI Assistants page (Connections → AI Assistants) shows every OAuth and manual token, plus active data connections created the first time an assistant successfully reads a source. Click Revoke to disable a connection immediately.
Every token has rolling call/data limits on top of your plan's exporter and rows limits. If something fails, the assistant gets a plain-language explanation, for example:
integration_required — the token doesn't have access to that source, or the underlying account needs reconnecting.scope_required — that data source wasn't included when the connection was authorized.plan_limit_reached — adding this AI connection would exceed your plan's exporter limit.budget_exhausted — a rolling call/data limit was hit; try again later or ask a narrower question.