Documentation

The MCP server

This page as Markdown, or in an AI chatMore ways to read this page

tidekeeper's figures, and how each was counted, for an AI agent. It only reads: nothing an agent does through it changes your sites, your goals or anything else.

# The MCP server

> tidekeeper's documentation: https://tidekeeper.eu/docs/mcp

tidekeeper's figures, and how each was counted, for an AI agent. It only reads: nothing an agent does through it changes your sites, your goals or anything else.

## Connect from anywhere

The server answers at `/api/v1/mcp` on this tidekeeper's address, over HTTP. An assistant that signs in through OAuth needs nothing else: it opens a page here where you let it in, and gets a token that only reads. With Claude Code:

```sh
claude mcp add --transport http tidekeeper https://tidekeeper.eu/api/v1/mcp
```

## Or with a token of your own

Make one under Settings, API tokens, that may read figures, for the sites the agent may see. It is all the agent will ever be allowed, however it signs in: it can read no visitor's or consent record, no data request, nothing of the plan or the members. Then send it in a header:

```sh
claude mcp add --transport http tidekeeper https://tidekeeper.eu/api/v1/mcp \
  --header "Authorization: Bearer $TIDEKEEPER_TOKEN"
```

With a client that reads a JSON file of servers:

```json
{ "mcpServers": { "tidekeeper": {
    "url": "https://tidekeeper.eu/api/v1/mcp",
    "headers": { "Authorization": "Bearer <your token>" } } } }
```

## Or from the command line

Where [the command line](https://tidekeeper.eu/docs/cli.md) is set up with this tidekeeper and a token, `tidekeeper mcp` is the same server on standard input and output:

```sh
claude mcp add tidekeeper -- tidekeeper mcp
```

## What it can do

- `sites`: The sites the token can read, each with its mode and whether hits arrive
- `status`: Every site at a glance, and which need attention
- `stats`: Any figures: metrics over a period, broken down and filtered
- `top`: The top pages, sources, countries, devices…
- `why`: How a figure was counted, and what was left out
- `changed`: What moved this period against the last
- `live`: Who is on a site now, by page
- `page`: One page: its views and visitors, and the clicks and scrolling on its stored copy
- `goals`: The goals, and how many converted
- `install`: The line for a site's pages, the proxy lines, and whether hits arrive

Every answer is the API's own JSON, with a line saying what it is. When a figure looks wrong, an agent can ask `why` and get the same step-by-step answer as the dashboard.

Connect from anywhere

The server answers at /api/v1/mcp on this tidekeeper's address, over HTTP. An assistant that signs in through OAuth needs nothing else: it opens a page here where you let it in, and gets a token that only reads. With Claude Code:

claude mcp add --transport http tidekeeper https://tidekeeper.eu/api/v1/mcp

Or with a token of your own

Make one under Settings, API tokens, that may read figures, for the sites the agent may see. It is all the agent will ever be allowed, however it signs in: it can read no visitor's or consent record, no data request, nothing of the plan or the members. Then send it in a header:

claude mcp add --transport http tidekeeper https://tidekeeper.eu/api/v1/mcp \
  --header "Authorization: Bearer $TIDEKEEPER_TOKEN"

With a client that reads a JSON file of servers:

{ "mcpServers": { "tidekeeper": {
    "url": "https://tidekeeper.eu/api/v1/mcp",
    "headers": { "Authorization": "Bearer <your token>" } } } }

Or from the command line

Where the command line is set up with this tidekeeper and a token, tidekeeper mcp is the same server on standard input and output:

claude mcp add tidekeeper -- tidekeeper mcp

What it can do

sites: The sites the token can read, each with its mode and whether hits arrive
status: Every site at a glance, and which need attention
stats: Any figures: metrics over a period, broken down and filtered
top: The top pages, sources, countries, devices…
why: How a figure was counted, and what was left out
changed: What moved this period against the last
live: Who is on a site now, by page
page: One page: its views and visitors, and the clicks and scrolling on its stored copy
goals: The goals, and how many converted
install: The line for a site's pages, the proxy lines, and whether hits arrive

Every answer is the API's own JSON, with a line saying what it is. When a figure looks wrong, an agent can ask why and get the same step-by-step answer as the dashboard.