---
updatedAt: 2026-10-07T09:15:26.000Z
agentTools:
projectIndex: https://developers.tidio.com/llms.txt
---
# Analytics API MCP
# Connect the Analytics API with MCP
MCP (Model Context Protocol) is an open standard that lets AI tools like Claude and Cursor pull live data from other services. Once you connect Tidio's Analytics API over MCP, you can ask your AI assistant questions like *"How many tickets did we close last week?"* or *"Show me Lyro's resolution rate by day"* and get answers straight from your Tidio data.
This guide walks you through connecting Tidio to **Claude Code**, **Claude Desktop**, **Cursor **and** ChatGPT Desktop**. Pick the tool you use and follow its section below.
## Step 1: Get your API credentials
Credentials are generated exactly as for the OpenAPI - see [instructions](https://developers.tidio.com/docs/openapi-authorization) to generate Client ID and Client Secret you'll use to authenticate.
## Step 2: Connect your AI tool
The Analytics MCP server lives at:
```
https://taa-mcp.data.tidio.com/mcp
```
Every tool below connects to this same address. They just ask for it in different ways.
### Claude Code
Open a terminal (a plain text window for typing commands, already built into your computer):
* **Mac:** press `Cmd + Space`, type `Terminal`, press Enter.
* **Windows:** press the Windows key (⊞), type `PowerShell`, press Enter.
Then paste one of the commands below, replacing the placeholder values `YOUR_CLIENT_ID` and `YOUR_CLIENT_SECRET` with your real credentials. with your real credentials.
**💻 Paste into your terminal — two-header method:**
```
claude mcp add --header "X-Tidio-Openapi-Client-Id: YOUR_CLIENT_ID" --header "X-Tidio-Openapi-Client-Secret: YOUR_CLIENT_SECRET" --transport http Tidio-Analytics-API https://taa-mcp.data.tidio.com/mcp
```
If you get `command not found: claude` when you try the command below, Claude Code isn't installed on your computer yet. Install it first following these [instructions](https://code.claude.com/docs/en/quickstart).
### Claude Desktop
Claude Desktop's settings UI doesn't yet support custom authentication headers, so we connect through a small, official bridge tool called `mcp-remote`. It runs locally on your machine and forwards your credentials securely. You only need Node.js installed (most computers already have it, or you can get it from [nodejs.org](https://nodejs.org)).
1. Open Claude Desktop and go to **Settings > Developer > Edit Config**. This opens (or creates) your `claude_desktop_config.json` file in a text editor, not a terminal.
2. Add the following, replacing the placeholder values `YOUR_CLIENT_ID` and `YOUR_CLIENT_SECRET` with your real credentials.
```json
{
"mcpServers": {
"tidio-analytics": {
"command": "npx",
"args": [
"mcp-remote",
"https://taa-mcp.data.tidio.com/mcp",
"--header",
"X-Tidio-Openapi-Client-Id:${TIDIO_CLIENT_ID}",
"--header",
"X-Tidio-Openapi-Client-Secret:${TIDIO_CLIENT_SECRET}"
],
"env": {
"TIDIO_CLIENT_ID": "YOUR_CLIENT_ID",
"TIDIO_CLIENT_SECRET": "YOUR_CLIENT_SECRET"
}
}
}
}
```
3. Save the file and restart Claude Desktop.
### Cursor
1. Follow Cursor instructions.
2. In the configuration file, add the following, replacing the placeholder values `YOUR_CLIENT_ID` and `YOUR_CLIENT_SECRET` with your real credentials.
```json
{
"mcpServers": {
"tidio-analytics": {
"url": "https://taa-mcp.data.tidio.com/mcp",
"headers": {
"X-Tidio-Openapi-Client-Id": "YOUR_CLIENT_ID",
"X-Tidio-Openapi-Client-Secret": "YOUR_CLIENT_SECRET"
}
}
}
}
```
3. Open Cursor's **Customize** panel in the sidebar to confirm "tidio-analytics" is listed and enabled.
### ChatGPT Desktop
1. Open ChatGPT and switch to the Codex / Work tab
2. Paste this prompt: Set up Tidio Analytics MCP in my local Codex configuration.
Server name: tidio-analytics
URL:
Prepare placeholders for these HTTP headers:
X-Tidio-Openapi-Client-Id
X-Tidio-Openapi-Client-Secret
Preserve existing settings. Open the configuration file and show me where to enter my credentials
3. Enter your Client ID and Client Secret in the configuration file and save it. **Keep your credentials out of the chat.**
4. Go to Settings → MCP servers → Restart and start a new Codex / Work chat.
If you face any issues, ask GPT to solve them for you.
## Optional: Build a single-header token
Some tools require single-header token. Follow these instructions to combine your Client ID and Client Secret into one value. It needs a one-line command, which you paste into a **terminal**:
* **Mac:** press `Cmd + Space`, type `Terminal`, press Enter.
* **Windows:** press the Windows key (⊞), type `PowerShell`, press Enter.
With that window open, copy the command for your system below, paste it in, and replace `YOUR_CLIENT_ID` and `YOUR_CLIENT_SECRET` with your real credentials from Step 1, then press Enter:
**💻 Paste into your terminal (Mac/Linux):**
```bash
auth_header="$(echo -n "YOUR_CLIENT_ID:YOUR_CLIENT_SECRET" | base64)"
echo $auth_header
```
**💻 Paste into your terminal (Windows PowerShell):**
```powershell
$auth_header = [Convert]::ToBase64String([Text.Encoding]::UTF8.GetBytes("YOUR_CLIENT_ID:YOUR_CLIENT_SECRET"))
$auth_header
```
The terminal prints a long string of letters and numbers right under your command, that's your token. Copy it, you'll paste it into your tool's config as one header: `Authorization: Bearer `.
## What you can ask for
Once connected, every Analytics API endpoint is available as a matching tool, one to one. See [Analytics API referece](https://developers.tidio.com/reference/get_agent-details) for details. Your AI assistant decides which tool(s) to call based on your question, so you don't need to know the exact names, but it's recommended to read the documentation to know what data is in scope.
## Good to know
* **Read-only.** These tools can only read your analytics data. They can't send messages, change settings, or modify anything in your account.
* **Same data, same limits.** The Analytics API pulls from the same dataset as the OpenAPI, follows the same [rate limits](https://developers.tidio.com/docs/rate-limiting) and [permissions](https://developers.tidio.com/docs/enable-analytics-api#endpoint-availability-by-plan:~:text=the%20OpenAPI.-,Endpoint%20availability%20by%20plan,-The%20Analytics%20API) and the same rules apply to [data freshness](https://developers.tidio.com/docs/data-availability).
* **One set of credentials, every tool.** You don't need separate keys for Claude, Cursor, or anything else. If you ever need to revoke access, disabling the Client Secret in **Settings > Developer > OpenAPI** disconnects it everywhere at once.
* **Large date ranges may need follow-up questions.** The underlying endpoints return up to 100 rows per call. Your AI assistant will usually page through results automatically, but if a question covers a lot of data, asking a narrower question (by date range or agent, for example) gets you a faster, more reliable answer.
## Prompts you can try
Once you're connected, just ask in plain English. A few to get you started:
1. What are the top reasons people contact us, such as shipping, returns or sizing. Analyze last 30 days.
2. When do we miss the most chats, by day of week and hour? This can help with staffing. Analyze last 30 days.
3. Which topics most often get handed from Lyro to a human agent? Analyze last 30 days.
4. Analyze chat ratings from the last 30 days. What do we need to focus on?
## Recreate a sample report with your own data
You can turn this [sample report](https://assets.tidio.com/analytics_api_docs/analytics_api_sample_report.html) into a report built on your own Tidio data, using an AI agent such as Claude.
### Step 1: Download the report
1. Open the download link. If the report opens in your browser, right-click the link and choose **Save link as…**.
2. Save the file as `analytics_api_sample_report.html` and keep the `.html` extension.
### Step 2: Connect your AI agent to Tidio Analytics
Make sure your AI agent has access to Tidio data through MCP, following the instrucitons from [here.](<# Connect the Analytics API with MCP>)
### Step 3: Upload the report to the agent and ask for you data
Start a new conversation and attach the downloaded file. Paste this prompt into the chat and send it together with the file:
> Recreate the attached report. Keep the same layout, sections and charts, but replace all sample data with my data from Tidio Analytics for the last 30 full days (ending yesterday). Use the Tidio Analytics tools to fetch the data. If a metric isn't available or a query is too large, use a smaller sample and say so in the report. Give me the result as a single HTML file.
### Step 4: Review and save
The agent returns an HTML file you can open in any browser. You can ask for changes in the same conversation, for example "show only agent Anna" or "add a chart of tickets per day".
### Tips
* On high-traffic accounts, fetching all conversations can take a long time. If it keeps failing, ask for a shorter period, such as 7 days.