MCP Server Docs
Query your restaurant analytics, reviews, and manager logs using natural language through any MCP-compatible AI assistant.
The Tenzo MCP server gives AI assistants access to your restaurant data through seven tools:
Sales, labour cost, covers, transactions — grouped by location, day, or category.
Customer reviews from Google, TripAdvisor, and other platforms, with ratings and text.
Manager logs and shift notes, filterable by location and date range.
The AI also calls these helper tools automatically at the start of each conversation:
get_analytics_context (Analytics Context) — Your currency, timezone, week start day, and date periods.get_available_metrics (Available Metrics) — Which metrics can be queried — e.g. net sales, labour cost.get_available_dimensions (Available Dimensions) — How results can be grouped — e.g. by location, day.get_available_filters (Available Filters) — How results can be filtered — e.g. by location, category.Read-only data. It cannot modify your data, only your Tenzo configuration.
Choose your AI client below. You'll be asked to sign in to your Tenzo account the first time you connect.
Tenzo is a custom connector. If you don't see it in your list, ask an admin to add it from claude.ai/admin-settings/connectors, pointing at https://ai.gotenzo.com/mcp.
Navigate to claude.ai/customize/connectors.
Find Tenzo in the list and click Connect.
You'll be redirected to sign in with your Tenzo account. Once authorized, the connector will appear as active in your settings.
Open a new conversation on claude.ai and ask about your restaurant data.
Tenzo is a custom connector. If you don't see it in your list, ask an admin to add it from claude.ai/admin-settings/connectors, pointing at https://ai.gotenzo.com/mcp.
In Claude Desktop, open Settings and navigate to the Connectors section.
Find Tenzo in the list and click Connect.
A browser window will open for you to sign in with your Tenzo account. Once authorized, return to Claude Desktop.
Open a new conversation and ask about your restaurant data. You should see the Tenzo tools available in the tools menu.
Run the following command in your terminal:
claude mcp add --transport http tenzo https://ai.gotenzo.com/mcp/Start a new conversation and ask something like "What were my sales last week?". You'll be prompted to authorize via OAuth on first use.
We've applied for a public ChatGPT App listing. In the meantime, you can use our CustomGPT.
Visit the Tenzo CustomGPT and start a conversation.
You'll be prompted to connect your Tenzo account on first use.
Once the public ChatGPT App is approved, we'll update these instructions with a simpler setup process.
Edit ~/.gemini/settings.json (create it if it doesn't exist).
{
"mcpServers": {
"tenzo": {
"url": "https://ai.gotenzo.com/mcp"
}
}
}Launch the Gemini CLI and ask a question about your restaurants. You'll be prompted to authorize via OAuth on first use.
Any MCP-compatible client can connect to the Tenzo server using:
| Setting | Value |
|---|---|
| Server URL | https://ai.gotenzo.com/mcp/ |
| Transport | Streamable HTTP |
| Authentication | OAuth 2.0 (handled automatically) |
If your client doesn't support remote MCP servers natively, use the mcp-remote bridge:
npx -y mcp-remote https://ai.gotenzo.com/mcp/ --transport http-firstIf your organization doesn't have the Tenzo connector yet, an admin can add it in a few steps:
Navigate to claude.ai/admin-settings/connectors.
Scroll to the bottom and click Add Custom Connector.
Type Tenzo as the name and enter the following as the Remote MCP server URL:
https://ai.gotenzo.com/mcpClick Add. The Tenzo connector will now be available to all members of your organization in their Connectors settings.
The Tenzo MCP server uses OAuth 2.0. The first time you connect from a new client, you'll be redirected to Tenzo's login page to sign in. After that, your session is cached and you won't need to sign in again until the token expires.
The MCP server only has access to data your Tenzo account can already see — it respects your existing permissions and business access.
Once connected, try asking your AI assistant things like:
| Category | Example |
|---|---|
| Sales | “Show me the revenue impact if my bottom 3 locations matched the upselling rate of my top performer.” |
| Labour | “On my highest COL % days, what did my GMs log? Is there a pattern in the reasons given?” |
| Inventory | “What's my waste cost per location this week, and which sites are outliers?” |
| Reviews | “What are the 5 most common negative themes in my reviews over the last 90 days, and which locations appear most frequently?” |
| Combined | “On my worst-reviewed days, what did labour, covers, and average spend per head look like vs my best-reviewed days?” |
Your session has expired or the initial OAuth flow didn't complete. Try:
Tenzo is added as a custom connector — it doesn't appear in Claude's built-in directory. If you don't see it, an admin needs to add it. See Adding a connector in Claude.
claude mcp listclaude mcp remove tenzo then re-add.curl https://ai.gotenzo.com/ should return a JSON status response.~/.gemini/settings.json is valid JSON.curl https://ai.gotenzo.com/If the connector is already authorised but tool calls are erroring or returning the wrong data, try these in order:
get_analytics_context.Your Tenzo account doesn't have access to the requested data. Check with your Tenzo administrator that your account has the correct permissions and business access.
The MCP server returns data from Tenzo's Public API, which reflects what you see in the Tenzo dashboard. If numbers don't match your expectations:
If you're still having trouble, contact Tenzo support at support@gotenzo.com with:
• Which AI client you're using (Claude, ChatGPT, Gemini, etc.)
• The error message or unexpected behaviour
• Your operating system
• The specific API request that was made (see below)
When submitting a support request, it is very helpful if you include the specific API call that was made. Ask your AI client something like "please show me the API call you just made" — it will show the tool name, parameters, and raw response.
